Parrot Anafi Ai. the First Serial Robotic 4G UAV

Introduction

Par­rot Anafi Ai is a com­plete­ly new unique solu­tion from a well-known French devel­op­er, adapt­ed pri­mar­i­ly for com­mer­cial use both in man­u­al and ful­ly auto­mat­ic modes in var­i­ous fields of human activ­i­ty, includ­ing: sur­vey of build­ings, infra­struc­ture, telecom­mu­ni­ca­tions tow­ers, wind tur­bines, solar bat­ter­ies, pipelines, refiner­ies, as well as in map­ping, geo­desy, 3D mod­el­ing, orthopho­tog­ra­phy, con­struc­tion mon­i­tor­ing, auto­mat­ed sur­veil­lance, real estate, insur­ance, video con­tent cre­ation.

The busi­ness card of Anafi Ai is the built-in 4G mod­ule, thanks to which the drone got the oppor­tu­ni­ty to use the fourth gen­er­a­tion of mobile com­mu­ni­ca­tions as the main data trans­mis­sion chan­nel between the drone and the con­trol equip­ment, which actu­al­ly removes all restric­tions on the flight range and pre­de­ter­mines the oper­a­tion of the UAV beyond the line of sight. Look­ing ahead, we note that the declared poten­tial of the car turned out to be incred­i­bly pow­er­ful and, more than ever, cor­re­sponds to its futur­is­tic appear­ance, the cre­ation of which French engi­neers were inspired by nature.

Developer about car

Mar­tin Liné, Mar­ket­ing Direc­tor at Par­rot, on how the ANAFI Ai drone was cre­at­ed: “At Par­rot, we are dri­ven by one pas­sion: bio­mimicry. Up to this point, Par­rot drones did not offer an obsta­cle avoid­ance fea­ture. But we thought about it for a long time and tried to find an ele­gant solu­tion to achieve this with­out hav­ing to mul­ti­ply the cam­eras around the drone, because it sim­ply com­pli­cates the hard­ware. We noticed that there are no known species in nature that have a sys­tem of vision dis­trib­uted through­out the body, but this does not pre­vent them from visu­al­iz­ing the world around them in all direc­tions. This is how we cre­at­ed ANAFI Ai.”

Peculiarities

Key fea­tures of the Par­rot Anafi AI:

• Drone flight weight: 898 grams
• Ready to fly time 60 sec­onds
• Unique portable fold­able design
• Built-in 4G com­mu­ni­ca­tion mod­ule for full drone/control equip­ment inter­ac­tion with­out restric­tions in any envi­ron­men­tal con­di­tions
• Com­pat­i­bil­i­ty with LTE world­wide
• Equipped with an omni­di­rec­tion­al sig­nal trans­mis­sion sys­tem (4 direc­tion­al reflec­tor anten­nas (with gain: 2.5 dBi / s anten­nas). The air­craft deter­mines the best pair of anten­nas depend­ing on its ori­en­ta­tion and posi­tion rel­a­tive to the posi­tion of the pilot)
• Equipped with a unique omni­di­rec­tion­al visu­al­iza­tion sys­tem of the sur­round­ing world built on a pair of stereo­scop­ic cam­eras mount­ed on a spe­cial­ized gim­bal with a 330-degree rota­tion range (from ‑120° to + 210°), work­ing in tan­dem with an intel­li­gent flight path opti­miza­tion sys­tem
• Adapt­ed for indoor flight
• Live video stream­ing in 1080p qual­i­ty with low laten­cy (300 m/s) with no lim­its on flight range any­where in the world, with a video bit rate of 12 Mbps
• Abil­i­ty to trans­fer files to cloud stor­age direct­ly in flight
• Built-in Wi-Fi com­mu­ni­ca­tion mod­ule (Wi-Fi 802.11a/b/g/n) for full inter­ac­tion of the drone / con­trol equip­ment (as an alter­na­tive con­nec­tion option)
• Slave. WiFi fre­quen­cy band: 2.4GHz/5.8GHz
• Max. Wi-Fi FPV Range: 9km
• The unique auto­mat­ic mode of oper­a­tion of 4G / Wi-Fi mod­ules allows, based on net­work con­ges­tion, to seam­less­ly switch from one com­mu­ni­ca­tion tech­nol­o­gy to anoth­er with­out data loss
• Drone pro­tec­tion class IPX3
• Max. flight time 32 min
• Max. hor­i­zon­tal flight speed 61 km/h
• Max. wind resis­tance: 13 m/s
• Max. ceil­ing above sea lev­el 5000 m
• 48MP 4K cam­era built on a 1/2‑inch CMOS sen­sor with a pix­el size of 0.8μm; with built-in Bay­er col­or fil­ter array
• Lens Opti­cal LD-ASPH
• Hybrid 6‑Axis Image Sta­bi­liz­er (3‑Axis Mechan­i­cal + 3‑Axis Dig­i­tal EIS)
• Con­trolled gim­bal range ‑90° to +90°
• 6x loss­less dig­i­tal zoom (2x in 4K mode/4x in FHD mode)
• Max. video bitrate 200 Mbps
• HDR8/HDR10 video record­ing for­mats
• Max. dynam­ic range 14EV
• P‑log col­or pro­file for post-pro­cess­ing
• 5 × auto­mat­ic flight and shoot­ing modes (Photogrammetry/Flight Plan/Cameraman/Smart RTH/Vehicle)
• Built-in slots for microSD and SIM cards
• Com­pat­i­ble with PIX4D Suite
• Pho­togram­met­ric flight modes are avail­able in mob. FreeFlight 7 app and Open­Flight soft­ware
• Abil­i­ty to cre­ate a flight plan in one click
• 48MP pho­togram­me­try at 1fps: twice as fast as the com­pe­ti­tion
• Shoot­ing accu­ra­cy: 0.46 cm/px. GSD at 30m
• Par­rot SDK sup­port (with the abil­i­ty to inject your own code into the drone using the Air SDK, cre­ate mobile appli­ca­tions using the Ground SDK, set up an open source ground sta­tion using Open­Flight, inter­act with the mod­ern Sphinx drone sim­u­la­tion tool, as well as the abil­i­ty to devel­op Python scripts using Olympe and video and meta­da­ta pro­cess­ing via PdrAW)
• Com­pat­i­ble with third par­ty soft­ware: PIX4Dscan, PIX4Dinspect, PIX4Dcloud, PIX4Dreact, PIX4Dsurvey
• ANAFI Ai com­plies with the Euro­pean Union Gen­er­al Data Pro­tec­tion Reg­u­la­tion (GDPR)
• High-End Data Pri­va­cy with Embed­ded NIST FIPS140‑2 Lev­el 3 Com­pli­ant WISeKey Secure Ele­ment with Com­mon Cri­te­ria EAL5+ Cer­ti­fi­ca­tion
• Video and con­trols are secured using SRTP/DTLS pro­to­cols in accor­dance with webRTC
• New portable con­trol equip­ment Par­rot Sky­con­troller 4
• New Par­rot FreeFlight 7 mobile app

Equipment

The devel­op­er offers the drone in a sin­gle con­fig­u­ra­tion, which will include:

1. Drone ANAFI AI
2. Hard trans­port case
3. Con­trol equip­ment Par­rot Sky­con­troller 4
4. 1×Smart bat­tery
5. USB charg­er
6. Spare Prop Kit (2CW/2CCW)
7. Com­pan­ion cable (USB‑C to Light­ning)
8. 2 × com­pan­ion cables (USB‑C to USB‑C)
9. Cal­i­bra­tion table

Specification

Drone

Mod­el:

Anafi AI

The weight:

898 g

Fold­ed size:

304×130×118mm

Unfold­ed size:

320×440×118mm

Prelaunch deploy­ment time:

60 sec

Max. flight time:

32 min

Max. hor­i­zon­tal flight speed:

61 km/h

Max. ver­ti­cal take­off speed:

14.5 km/h

Max. wind resis­tance:

12.7 m/s

Max. num­ber of pro­peller rev­o­lu­tions:

10,000 rpm

Max. ceil­ing above sea lev­el:

5 000 m

Slave. tem­per­a­ture range:

-10°C to 40°C

Pro­tec­tion class:

IPX3

Nois­i­ness:

82 dB at 1 m

Built-in slots:

MicroSD and SIM card

Connection

Con­nec­tion:

4G/Wi-Fi

Video stream qual­i­ty:

1920 × 1080 at 30 fps

Navigation system

Satel­lite nav­i­ga­tion:

GPS, Glonass and Galileo

Sen­sors:

Ver­ti­cal Cam­era
Barom­e­ter
Mag­ne­tome­ter
2 × 6‑axis iner­tial units
2 × 3‑axis accelerom­e­ters
2 × 3‑axis gyro­scopes
Stereo­scop­ic cam­eras mount­ed on a 311° gim­bal (-107° to +204°)

ON:

Intel­li­gent Flight Tra­jec­to­ry Opti­miza­tion Sys­tem

Skycontroller 4

Dimen­sions with­out smart device:

238×147×55mm

Max. the size:

315×147×55mm

The weight:

606 g

Oper­at­ing range fre­quen­cies:

2.4GHz — 5GHz

The qual­i­ty of the dis­played FPV pic­ture:

1080P

Bat­tery capac­i­ty:

3350 mAh 7.2V

Max. com­pat­i­ble device size:

8 inch­es

Ports:

2 × USB‑C (charg­ing + con­nec­tion), micro-HDMI

Pro­tec­tion class:

IP5X

Drone Intelligent Battery

Type of:

high den­si­ty lithi­um poly­mer (262 Wh/kg)

Capac­i­ty:

6800 mAh 4.4 V

Charg­ing port:

USB‑C

The weight:

366 g

Max. mem­o­ry pow­er:

45 W

App

Mob. Appen­dix:

Par­rot FreeFlight 7 (free; no fee)

Down­load resources:

App Store

Com­pat­i­ble OS:

iOS 12 and above

Options:

3 free 3D mod­els cour­tesy of PIX4Dcloud

Camera

Matrix:

48MP 1/2″ CMOS

Pix­el Size:

0.8 µm

Array of col­or fil­ters:

Quad bay­er array

Dynam­ic Range:

14EV in HDR mode

Lens:

Opti­cal LD-ASPH (Low Dis­per­sion Aspher­i­cal Lens): Aper­ture: f/2.0
Focal length in 35mm equiv­a­lent: 24mm

Depth of field:

from 4.5 m to ∞

ISO range:

from 50 to 6400

Excerpt:

1/15 to 1/10000 s

Zoom:

6x Loss­less Zoom:
up to 4x (in 1080p mode) and 2x (in 4K UHD mode)

Image Sta­bi­liza­tion:

6‑axis hybrid:
Mechan­i­cal: 3‑axis (Pitch, Roll, Yaw)
Elec­tron­ic (EIS): 3‑axis (Pitch, Roll, Yaw)

Slave. gim­bal tilt range:

-90° to + 90°

Depth of field:

from 4.5 m to ∞

Video for­mats:

MP4 (H.264, H.265)

Video modes:

4K UHD: 3840 × 2160 @ 24/25/30/48/50/60 fps
1080P: 1920 × 1080 @ 24/25/30/48/50/60/90/100/120 fps

HDR10:

4K UHD/1080p@24/25/30 fps

HDR8:

for all modes

Max. bitrate:

200 Mbps

Hor­i­zon­tal FOV (HFOV):

68°

Col­or pro­file:

P‑Log

Pho­to for­mats:

JPEG, DNG (Dig­i­tal Neg­a­tive RAW)

Per­mis­sion:

48MP (8000×6000), 12MP (4000×3000)

Hor­i­zon­tal FOV (HFOV):

73° (Wide), 69° (Wide with EIS), 65° (Straight with EIS)

Pho­to modes:

Sin­gle, Brack­et­ing, Burst (10 fps), Panora­ma (4 for­mats), Time-lapse, GPS-lapse

Autonomous flight modes

Pho­togram­me­try:

Sin­gle grid/Double grid/Orbit

flight plan:

Mul­ti­ple Waypoints/Points Of Inter­est

cam­era­man:

Auto­mat­ic fram­ing with visu­al track­ing

Smart RTH:

auto return to take­off point with adjustable flight alti­tude

Vehi­cle:

the flight is tied to the con­troller that sets the loca­tion (allows you to land on mov­ing vehi­cles)

cyber security

Pecu­liar­i­ties:

Zero data is trans­mit­ted with­out user con­sent
FIPS140‑2 com­pli­ant and CC EAL5+ secu­ri­ty fea­ture cer­ti­fi­ca­tion
Strong authen­ti­ca­tion for 4G
Pho­tos dig­i­tal­ly signed
Trans­paren­cy and con­tin­u­ous secu­ri­ty check of Bug Boun­ty

Parrot SDK

Air SDK:

built-in encod­ing capa­bil­i­ties

Ground SDK:

iOS and Android app devel­op­ment kit

open­flight:

Free-Flight 7 open source core

Sphinx:

3D pho­to­re­al­is­tic sim­u­la­tor

Olympic:

Python con­troller pro­gram­ming inter­face

PDRAW:

video and meta­da­ta toolk­it

Below you can read the Par­rot Anafi Ai key infor­ma­tion, which will allow you to learn more about the declared poten­tial of the machine, as well as what improve­ments have been made com­pared to the com­mer­cial ver­sion of the Par­rot Anfi Ther­mal drone and oth­er drones from lead­ing brands.

Appearance

Connection

4G

ANAFI Ai includes a 4G radio mod­ule (in addi­tion to a Wi-Fi radio mod­ule) that allows you to broad­cast a video stream in 1080p qual­i­ty with a max­i­mum bit rate of 12 Mbps, with very low laten­cy (300 ms) with­out range lim­i­ta­tion and any­where in the world.

Compatibility

The ANAFI Ai 4G mod­ule sup­ports over 28 fre­quen­cy bands cov­er­ing over 98% of the fre­quen­cies deployed world­wide.

Automatic network switch

The qual­i­ty and through­put of 4G and Wi-Fi net­works are mea­sured every 100ms to adapt stream­ing to net­work con­di­tions. Com­bined with rout­ing algo­rithms, the con­nec­tion between the air­craft and its con­troller is main­tained even dur­ing severe Wi-Fi dis­rup­tion. So when the effec­tive band­width (use­ful band­width) of Wi-Fi is below 1.5 Mbps, the sys­tem will auto­mat­i­cal­ly switch to 4G.

To lim­it the con­sump­tion of mobile data when the pilot is with­in range of the air­craft’s Wi-Fi net­work, the tran­si­tion from 4G to Wi-Fi is also per­formed auto­mat­i­cal­ly with­out clip­ping the video stream.

Max. range with 4G connection

Devel­op­er’s video shows max. flight range of Anafi Ai with 4G con­nec­tion in one direc­tion on one bat­tery charge. The total flight time was 29 min­utes at an aver­age hor­i­zon­tal flight speed of 58 km/h (16 m/s). Dur­ing this time, the drone was able to cov­er 27 km. You can also notice that the flight was car­ried out in strong wind con­di­tions, which the drone sys­tem con­stant­ly warned about.

Implemented video stream optimization algorithms

4G congestion control

The over­load con­trol algo­rithm allows:

• Mea­sure pack­et loss over the entire net­work cycle.
• Mea­sure laten­cy (round trip time).
• Adjust the band­width accord­ing to these two para­me­ters.

The ulti­mate goal of the algo­rithm is to max­i­mize the avail­able band­width while main­tain­ing the low­est pos­si­ble laten­cy. This algo­rithm is imple­ment­ed on each of the inter­faces avail­able in the drone, each with its own para­me­ters opti­mized accord­ing to the net­work. Based on the infor­ma­tion pro­vid­ed by this algo­rithm, the link man­ag­er decides on the rout­ing and active inter­face.

Drone control via 4G connection

ANAFI Ai con­nects to the remote con­troller via 4G in less than 30 sec­onds if the air­craft is out of Wi-Fi range and in less than 15 sec­onds if the air­craft is in Wi-Fi range. And also 4G con­nec­tion pro­vides:

• Con­nec­tion dis­cov­ery and ini­ti­a­tion based on the VOIP SIP pro­to­col.
• Using a relay serv­er to estab­lish a con­nec­tion in secure net­works.

Video stream performance

• Laten­cy: 300ms.
• Secu­ri­ty: Videos and con­trols are secured using SRTP/DTLS in accor­dance with webRTC.
• Anten­nas: 28 LTE bands from 700 MHz to 2.6 GHz.

WiFi

Omnidirectional transmission system

• ANAFI Ai has 4 direc­tion­al reflec­tor anten­nas (gain: 2.5dBi/s anten­na). The drone deter­mines the best pair of anten­nas based on its ori­en­ta­tion and posi­tion rel­a­tive to the pilot’s posi­tion.
• With a recom­bined gain of 3.5 +/- 1.5 dBi in the hor­i­zon­tal plane of the drone, the gain of the ANAFI Ai radio sig­nal is high­ly uni­form.
• Down­ward radi­a­tion from the anten­nas has been improved by +4 dB com­pared to ANAFI.

High Power Radio External Design

The front of the radio allows pow­er to be max­i­mized at the base of the anten­na with a very good lev­el of lin­ear­i­ty and sen­si­tiv­i­ty (-94 dBm at 6.5 Mbps) to reach the max­i­mum FCC pow­er lim­it.

Reliability of Wi-Fi connection

802.11 options

A sub­set of the pro­to­col para­me­ters was cho­sen to opti­mize per­for­mance in drone envi­ron­ments such as rel­a­tive­ly low through­put, low laten­cy, vari­abil­i­ty in recep­tion due to air­craft speed, long range, and inter­fer­ence. These para­me­ters include aggre­ga­tion, retries, MiMo tech­nol­o­gy (STBC), con­trol frame datarate, and dis­con­nect con­di­tions.

Smart Interference Prevention System

ANAFI Ai has an algo­rithm for elim­i­nat­ing chan­nels (in two bands 2.4 GHz and 5 GHz) in case of inter­fer­ence.

Adaptation and flow monitoring

ANAFI Ai con­stant­ly mon­i­tors the sta­tus of its con­nec­tion at a fre­quen­cy of 4 Hz and can detect the pres­ence of inter­fer­ence. This allows you to dynam­i­cal­ly opti­mize the through­put and size of trans­mit­ted pack­ets. It also warns the pilot if he is in a par­tic­u­lar­ly noisy envi­ron­ment or close to los­ing sig­nal.

Bandwidth reduction

At the lim­it of its range, and if con­di­tions allow it, ANAFI Ai can switch to 10 MHz band­width to improve its sen­si­tiv­i­ty by 3 dB and increase its range by 40%.

Radio indicators

Video broadcast

Video link metrics

Implemented video stream optimization algorithms

Parrot Gen4 Streaming (4th generation)

This algo­rithm reduces the visu­al impact of net­work loss and allows all decoders to com­mu­ni­cate while pro­vid­ing a syn­tac­ti­cal­ly com­plete stream: miss­ing parts of the image are recon­struct­ed as miss­ing parts that are iden­ti­cal to those in the ref­er­ence image.

Thus, glitch­es are con­tained only in those areas that are prone to loss, and do not apply to the entire image.

The graphs below show the mac­roblock decod­ing suc­cess rates at 5% net­work loss, with and with­out advanced ANAFI Ai stream­ing fea­tures. The algo­rithm pro­vides cor­rect decod­ing of 75% of mac­roblocks. They allow the user to con­tin­ue their mis­sion with­out screen freezes or loss of stream­ing.

Overload control

The algo­rithm also eval­u­ates the Wi-Fi and radio envi­ron­ment to antic­i­pate and avoid pack­et loss and net­work con­ges­tion, which helps reduce laten­cy. The algo­rithm is based on an esti­mate of the chan­nel through­put cal­cu­lat­ed from the data rate and the error rate at the phys­i­cal lay­er; it then affects the net­work encod­ing and encap­su­la­tion para­me­ters.

metadata

Meta­da­ta is trans­mit­ted with the video stream. In par­tic­u­lar, they con­tain ele­ments of drone teleme­try (posi­tion, alti­tude, speed, bat­tery lev­el, etc.) and video met­rics (cam­era angle, expo­sure val­ue, field of view, etc.).

Syn­chro­niza­tion of images and open meta­da­ta per­forms the func­tions of pre­cise posi­tion­ing on the map, track­ing flight instru­ments in the HUD, or enabling aug­ment­ed real­i­ty ele­ments.

Inclu­sion of meta­da­ta is car­ried out by stan­dard meth­ods (RTP head­er exten­sion); the data for­mat defined by Par­rot is open: it is avail­able in the ANAFI Ai SDK.

Camera

48MP matrix

The ANAFI Ai sen­sor includes a large num­ber of megapix­els to pro­vide detailed aer­i­al pho­tog­ra­phy.

It uses Quad Bay­er col­or fil­ter array tech­nol­o­gy, where groups of 4 adja­cent pix­els have the same col­or. Thus real-time HDR cap­ture can be obtained in both pho­to and video modes by adding sig­nals from four neigh­bor­ing pix­els.

Its dynam­ic range is 4 times greater than stan­dard Bay­er matri­ces. Even com­plex scenes can be cap­tured with min­i­mal high­light high­lights or loss of shad­ow detail.

Lens

The ANAFI Ai lens was spe­cial­ly designed for Par­rot. It com­bines 6 aspher­i­cal ele­ments and is opti­mized to reduce opti­cal glare. This lens deliv­ers 68° HFOV in stan­dard video mode and 64.6° HFOV in stan­dard pho­to mode.

Video Modes

ANAFI Ai shoots smooth 4K video at 60fps, includ­ing in P‑Log, as well as HDR10 4K video at up to 30fps. The table below lists all ANAFI Ai video modes.

Video encoding

Users can choose between H.264 (AVC) and H.265 (HEVC) for­mats.

The fol­low­ing pix­el for­mats are used for all res­o­lu­tions:

• YUV420p (8 bits/component, BT.709 col­or space) for stan­dard mode and HDR8.
• YUVJ420p (8 bits/component, full range — BT.709 col­or space) for P‑log style.
• YUV420p10 (10 bits/component, BT.2020 col­or space) for HDR10 record­ing, H.265 only.

HDR

When record­ing video in HDR8 and HDR10, ANAFI Ai cov­ers a dynam­ic range of 14EV. The HDR10 for­mat pro­vides a max­i­mum bright­ness of 1000 nits and a col­or depth of 10 bits. It pro­vides a bil­lion col­or palette ver­sus 16 mil­lion for the stan­dard dynam­ic range. Com­pared to HDR8, HDR10 deliv­ers images that are more than twice as bright with a cor­re­spond­ing increase in con­trast. HDR8 can be dis­played on any stan­dard screen, while HDR10 is for TVs and HDR10 screens.

Photo modes

Description of modes

The mode con­trol sec­tion con­tains set­tings that affect the num­ber of shots that are tak­en each time the shut­ter is released.

single capture mode

Stan­dard frame mode. After each shut­ter release, the cap­tured image is imme­di­ate­ly processed by the sys­tem.

bracketing mode

Users can shoot a series of 3, 5 or 7 frames with dif­fer­ent expo­sures for each frame. The fol­low­ing pre­sets are avail­able:

• [-1 EV, 0, +1 EV] (default set­tings)
• [-2 EV, ‑1 EV, 0, +1 EV, +2 EV]
• [-3 EV, ‑2 EV, ‑1 EV, 0, +1 EV, +2 EV, +3 EV]

burst mode

Burst mode will allow the user to take a series of 10 frames in 1 sec­ond.

panorama mode

Panora­ma mode includes four dif­fer­ent panora­ma shots:

1. Spher­i­cal (360°) — This panoram­ic mode includes three options for shoot­ing a spher­i­cal panora­ma: Sphere (Sphere) / Lit­tle Plan­et (Lit­tle Plan­et) / Tun­nel (Tun­nel).
2. Hor­i­zon­tal (180°)
3. Ver­ti­cal (109°)
4. super­wide — new ultra-wide mode that pro­vides stitch­ing of 9 images (HFOV 110°, rec­ti­lin­ear panora­ma)

See the table below for details on the char­ac­ter­is­tics of panoram­ic modes:

time-lapse mode

This mode allows you to take pic­tures at the fol­low­ing fixed inter­vals:

• 48Mp: 1, 2, 4, 10, 30 or 60s.
• 12Mp: 0.5, 1, 2, 4, 10, 30 or 60s

GPS lapse mode

This pho­to mode was devel­oped for inspec­tion and pho­togram­me­try. It allows you to take pic­tures at the fol­low­ing fixed dis­tance inter­vals: 5, 10, 20, 50, 100 or 200 meters.

The table below shows the shoot­ing modes and res­o­lu­tions, includ­ing the Sen­sor Read­out mode:

Settings

The fol­low­ing table lists the avail­able set­tings for each mode.

6x zoom

Zoom­ing is avail­able in all pho­to and video modes. Com­bined with a 48-megapix­el sen­sor, pre­cise sharp­en­ing algo­rithms pro­duce high-def­i­n­i­tion images even when using 6x dig­i­tal zoom. ANAFI Ai users can now see 1cm details from 75m away. The high pix­el count also allows 4K videos to be cropped to 1080p with­out qual­i­ty loss.

Hybrid stabilization

The ANAFI Ai cam­era has the most accu­rate sta­bi­liza­tion on the micro-UAV mar­ket.

It com­bines com­bined sta­bi­liza­tion:

• 3‑axis mechan­i­cal (3‑axis mech. gim­bal)
• 3 axis elec­tron­ic (EIS)

Mechan­i­cal sta­bi­liza­tion sta­bi­lizes the cam­er­a’s point­ing axis regard­less of the air­craft’s flight posi­tion. Elec­tron­ic image sta­bi­liza­tion cor­rects the effect of micro-vibra­tions for fre­quen­cies above 100 Hz, which can­not be han­dled by a mechan­i­cal dri­ve.

Main camera gimbal

Mechan­i­cal sta­bi­liza­tion allows you to sta­bi­lize and ori­ent the hor­i­zon­tal axis of view of the cam­era along all 3 axes.

3 axes of rota­tion mech. sus­pen­sion of the main cam­era ANAFI Ai.

Key Features

• 3 axis mechan­i­cal sta­bi­liza­tion for the main cam­era
• 292° ver­ti­cal off­set, ‑116° to +176° field of view

Gimbal performance

• The EIS algo­rithm cor­rects the effects of wide-angle lens wob­ble and dis­tor­tion, as well as 3‑axis dig­i­tal image sta­bi­liza­tion (Roll, Pitch and Yaw).
• The method con­sists in apply­ing a geo­met­ric trans­for­ma­tion of the image. The geom­e­try trans­for­ma­tion is linked to the time­stamp and pre­cise posi­tion thanks to the IMU.
• A geo­met­ric trans­for­ma­tion is applied to each image accord­ing to the opti­cal dis­tor­tions, oscil­la­tions and move­ments of the mea­sured cam­era mod­ule.

Swivel range 292°

The cam­era has gained a hor­i­zon­tal swiv­el range of ‑116°/ +176° around the Pitch axis, thus pro­vid­ing obser­va­tion above and below the drone, which is a unique oppor­tu­ni­ty in the micro-UAV mar­ket.

Photogrammetry

Aer­i­al pho­tog­ra­phy using unmanned aer­i­al vehi­cles is chang­ing the way spe­cial­ists con­duct inspec­tions and geo­det­ic works. Pho­togram­me­try meth­ods are used to process images col­lect­ed by UAVs to cre­ate 2D and 3D mod­els, which sub­se­quent­ly allow cus­tomers to plan the main­te­nance of sur­veyed objects in a time­ly man­ner.

4G pro­vides unpar­al­leled reli­a­bil­i­ty of the data trans­mis­sion chan­nel from the drone. Users can oper­ate the UAV in large areas, near met­al struc­tures, build­ings, with­out fear of loss of com­mu­ni­ca­tion.

Inspection and cartography

By com­bin­ing the AI ​​of the ANAFI Ai drone with the PIX4Dinspect online plat­form, users will be able to con­duct inspec­tions faster and more effi­cient­ly than ever. Machine learn­ing algo­rithms rec­og­nize anten­nas on cell tow­ers, deter­mine their size, height, tilt, azimuth and ver­ti­cal­i­ty.

Click on the images below to explore the 3D demo mod­els gen­er­at­ed by ANAFI Ai .

Example #1

Example #2

Best in class matrix

The ANAFI Ai drone cam­era is equipped with a 48MP 1/2‑inch CMOS sen­sor built with Quad Bay­er col­or fil­ter tech­nol­o­gy, which is more than ever suit­able for inspec­tion and pho­togram­me­try. The matrix involved allows you to get detailed images with a wide dynam­ic range.

High resolution photos

ANAFI Ai is capa­ble of gen­er­at­ing 48-megapix­el still images, there­by cap­tur­ing all the details in high res­o­lu­tion and cre­at­ing a high-den­si­ty point cloud.

exponential sharpness

Inspec­tion mis­sions require the abil­i­ty to iden­ti­fy minute details such as ser­i­al num­bers, con­nec­tors, rust spots and incip­i­ent cracks.

Wide dynamic range

10 stops of dynam­ic range in stan­dard mode, 14 stops in HDR mode. Opti­mal image gra­da­tion is essen­tial to cre­ate con­sis­tent cloud points and high qual­i­ty 2D or 3D recon­struc­tions.

55% more detail than 1″ sensors

The 48MP 1/2‑inch Quad Bay­er sen­sor of the ANAFI Ai drone is supe­ri­or in sharp­ness to the 20MP 1‑inch sen­sors that are used in a num­ber of mod­ern pro­fes­sion­al drones. The fol­low­ing images tak­en dur­ing the same height inspec­tion of the roof with ANAFI Ai and DJI Phan­tom 4 Pro V2.0 drones clear­ly demon­strate this fact.

Ideal for inspections

The ANAFI Ai gim­bal includes a 6‑axis hybrid (mechan­i­cal + elec­tron­ic) sta­bi­liza­tion sys­tem that com­pen­sates for flight fluc­tu­a­tions and guar­an­tees image sharp­ness. The ANAFI Ai cam­era has an adjustable range of ‑90° to +90°, mak­ing it the ide­al aer­i­al tool for view­ing the under­side of the bridge base.

Con­trolled tilt ± 90°

Shooting Accuracy

ANAFI Ai allows users to achieve a GSD of 0.46 cm/px from a height of 30 m, which means a rel­a­tive planime­try accu­ra­cy of up to 0.92 cm.

By com­par­i­son, at the same height, the DJI Phan­tom 4 Pro V2 only deliv­ers a GSD of 0.82cm/px. In oth­er words, the ANAFI Ai can map the same tar­get while fly­ing more than 1.5 times high­er than the Phan­tom at an equiv­a­lent lev­el of detail.

AI and 4G Capabilities

One application. Any flight plan

The poten­tial of the FreeFlight 7 mobile appli­ca­tion allows the user to launch all sur­vey, inspec­tion and pho­togram­me­try mis­sions.

Available photogrammetric flight modes

The fol­low­ing pho­togram­met­ric flight modes are now avail­able in the FreeFlight 7 mobile app:

Create a flight plan in one tap

One tap on the FreeFlight 3D inter­ac­tive map is all it takes to quick­ly scan a build­ing. Arti­fi­cial intel­li­gence auto­mat­i­cal­ly deter­mines the opti­mal flight para­me­ters and tra­jec­to­ry. 48MP images with pre­cise geo-ref­er­ence of ANAFI Ai sen­sors (IMU, GNSS and Time of Flight) enable accu­rate 3D recon­struc­tion.

Auto­mat­ic flight plan cre­at­ed with one tap in 3D land reg­istry.

• The map back­ground for the FreeFlight 7 appli­ca­tion is from ArcGIS soft­ware. The 3D build­ing rep­re­sen­ta­tion is based on Open­StreetMap data cov­er­ing cities around the world.
• Visu­al­iza­tion sys­tems ensure the safe­ty of a giv­en flight plan: users do not have to wor­ry about obsta­cles. ANAFI Ai autonomous­ly avoids them.

autonomous flight

Based on stereo­scop­ic vision, the ANAFI Ai omni­di­rec­tion­al sen­sor sys­tem auto­mat­i­cal­ly ori­ents itself in the direc­tion of trav­el.

The drone detects obsta­cles at a dis­tance of 30 m. As the flight plan is exe­cut­ed, the arti­fi­cial intel­li­gence tech­nol­o­gy con­stant­ly builds and updates the infill grid. It rep­re­sents the envi­ron­ment of the drone in vox­els.

Algo­rithms deter­mine the best tra­jec­to­ry to avoid obsta­cles while the drone remains focused on its goal: the intend­ed flight mis­sion.

4G connection

ANAFI Ai is the first com­mer­cial micro­drone equipped with a 4G mod­ule. It cov­ers over 98% of the fre­quen­cies used world­wide.

Inspec­tion of high-volt­age pow­er lines. Shot with ANAFI Ai, processed with Pix4Dmatic. Scene size: 4060 × 60 × 70 m. Num­ber of images: 2172. GSD: 1.3 cm/px. Height: 90 m. Front/side over­lap: 90%/65%.

The drone is able to seam­less­ly switch from Wi-Fi to 4G and vice ver­sa, thus pro­vid­ing the most reli­able con­nec­tion, which in turn guar­an­tees:

• Abil­i­ty to per­form BVLOS flights
• Sta­ble con­nec­tion, even when fly­ing in envi­ron­ments clut­tered with obsta­cles and build­ings
• Safe flight in high inter­fer­ence envi­ron­ments
• Video link qual­i­ty 1080p at 30 fps
• Direct upload of images to cloud servers

Flight time optimization

Accord­ing to the devel­op­er, sig­nif­i­cant time sav­ings have been achieved due to the high image per­for­mance of ANAFI Ai:

• 48MP allows the drone to fly more than 1.5 times high­er than drones with 20MP 1‑inch sen­sors while achiev­ing the same GSD. In oth­er words, high­er alti­tude and faster mis­sion go hand in hand.
• 1fps pho­tog­ra­phy: ANAFI Ai shoots twice as fast as Autel EVO 2 and DJI Phan­tom 4 Pro V2.0.

4G in-flight transmission on PIX4Dcloud

In the process of cre­at­ing a dig­i­tal mod­el from aer­i­al pho­tographs of a drone, file trans­fer and pho­to pro­cess­ing are two time-con­sum­ing tasks. ANAFI Ai helps users speed up their work­flow. In par­tic­u­lar, the drone sys­tem allows:

• Trans­fer images to secure servers dur­ing flight using the drone’s 4G con­nec­tion.
• Imme­di­ate­ly start com­put­ing objects at the end of the flight: ortho­mo­saics, point cloud, ele­va­tion mod­els and tex­tured mesh.
• Eas­i­ly share 2D maps and 3D sur­vey-qual­i­ty mod­els with employ­ees and cus­tomers.

Compatible with PIX4D Suite

ANAFI Ai is ful­ly com­pat­i­ble with a unique soft­ware suite of mobile, desk­top and cloud pho­togram­me­try appli­ca­tions, con­sist­ing of:

Photogrammetry or lidar

Why choose pho­togram­me­try for sur­veys and inspec­tions? In addi­tion to afford­abil­i­ty and ease of use, aer­i­al pho­togram­me­try is the best choice when visu­al inter­pre­ta­tion of data is required.

Photogrammetry Pros

• Pro­vides sev­er­al visu­al­iza­tions: ortho­mo­sa­ic, col­ored point cloud, tex­tured mesh.
• Cre­ates a high­er den­si­ty point cloud, each con­tain­ing a sig­nif­i­cant amount of infor­ma­tion (height, tex­ture, col­or).
• Out­per­forms LIDAR in accu­ra­cy of pho­to­re­al­is­tic 2D and 3D ren­der­ings — LIDAR does not pro­vide a high lev­el of con­tex­tu­al detail.

LIDAR cons

• It costs 100 times more than an RGB UAV.
• Com­plex­i­ty of data pro­cess­ing and pos­si­ble errors of inter­pre­ta­tion.
• Adding col­or infor­ma­tion is pos­si­ble, but it com­pli­cates the process.
• LIDAR only gen­er­ates a point cloud and is designed for spe­cif­ic cas­es and sit­u­a­tions.

SDK

The Par­rot Soft­ware Devel­op­ment Kit is a set of open source tools and soft­ware for devel­op­ers. Since the cre­ation of the first AR Drone, the com­pa­ny’s open source code has been at the heart of devel­op­er plat­forms and tools, and Par­rot has con­tin­u­ous­ly con­tributed to the open source devel­op­er com­mu­ni­ty.

The devel­op­er clar­i­fies that all the com­pa­ny’s SDK resources are avail­able free of charge, with­out reg­is­tra­tion and track­ing, on the Par­rot Devel­op­er Por­tal. You can also join thou­sands of devel­op­ers on the Par­rot forum where you can dis­cuss hot top­ics direct­ly with the com­pa­ny’s engi­neers.

Air SDK

Run your code on ANAFI Ai

The Air SDK pro­vides a rev­o­lu­tion­ary tech­nol­o­gy archi­tec­ture for down­load­ing and run­ning code direct­ly on the ANAFI Ai. Devel­op­ers can pro­gram tai­lor-made flight mis­sions with access to all drone sen­sors, con­nec­tiv­i­ty inter­faces, and autopi­lot fea­tures.

Air SDK pro­vides onboard access to:

• All sen­sors (IMU, GPS, TOF) and flight modes
• Video stream and meta­da­ta from all cam­eras
• Com­mu­ni­ca­tion inter­faces Wi-Fi, 4G, USB
• Depth maps and infill grids
• Cre­at­ing an obsta­cle avoid­ance tra­jec­to­ry

Any devel­op­er can:

• Change the state of the drone by cre­at­ing flight tasks
• Change nav­i­ga­tion mode
• Add embed­ded Lin­ux process­es (for exam­ple, trans­fer data over a 4G chan­nel or use com­put­er vision)

Air SDK sup­ports C++ or Python pro­gram­ming lan­guage. The Air SDK comes with a detailed instal­la­tion guide and API doc­u­men­ta­tion. Many appli­ca­tion exam­ples illus­trate all the pos­si­bil­i­ties offered by its unique archi­tec­ture.

Ground SDK

Build a powerful mobile app

The Ground SDK is a ground con­trol sta­tion (GCS) soft­ware plat­form for mobile devices (both iOS and Android are sup­port­ed). It allows any devel­op­er to cre­ate a mobile appli­ca­tion for ANAFI Ai to con­trol the drone direct­ly from a mobile device. All WAN func­tions (con­trol, video, set­tings) are avail­able through an easy to use and ful­ly doc­u­ment­ed API.

openflight

Open source ground control station

For the first time, Par­rot is open source for its ground con­trol sta­tion appli­ca­tion. Open­Flight is the core of our famous open source FreeFlight 7 appli­ca­tion. This allows the devel­op­er to focus on adding their own fea­tures and get a pro­fes­sion­al-look­ing app ready to be pub­lished to the App­Store right away.

Open Flight con­tains:

• All UX FreeFlight 7
• All inter­face set­tings
• Code for real-time 3D visu­al­iza­tion of obsta­cles
• 4G com­mu­ni­ca­tion con­trol code
• Open­Flight is pub­lished under a BSD‑3 license with a com­plete instal­la­tion guide and com­pre­hen­sive doc­u­men­ta­tion.

Sphinx

3D photorealistic modeling

Par­rot Sphinx is a state of the art drone sim­u­la­tion tool. Par­rot engi­neers use it to devel­op and test all ANAFI Ai fea­tures. The gen­er­al con­cept is to sim­u­late the actu­al firmware of a drone with all of its sen­sors in a visu­al­ly and phys­i­cal­ly real­is­tic envi­ron­ment.

Par­rot Sphinx allows you to:

• Sim­u­late all cam­eras and sen­sors
• Mod­el depth maps and image seg­men­ta­tion
• Nav­i­gate through many real­is­tic 3D scenes
• Con­nect to dif­fer­ent types of remote con­trols
• Use scripts to con­trol sim­u­la­tion
• Add pedes­tri­ans and vehi­cles
• Visu­al­ize and record flight data
• Cus­tomize drone sen­sors and sur­round­ing phys­i­cal ele­ments
• Con­trol the real time fac­tor

The Par­rot Sphinx is built with indus­try-lead­ing stan­dard com­po­nents:

Olympe

Python to control ANAFI Ai

Olympe pro­vides a Python con­troller pro­gram­ming inter­face for ANAFI Ai. The orig­i­nal pur­pose of Olympe is to inter­act with the Sphinx mod­el­ing envi­ron­ment. Olympe can also con­trol the phys­i­cal ANAFI Ai from a remote com­put­er.

The Olympe frame­work allows you to:

• Con­nect to a sim­u­lat­ed or phys­i­cal ANAFI Ai
• Send com­mand mes­sages (pilot­ing, cam­era ori­en­ta­tion, return home, flight plan)
• Start and stop stream­ing video from all cam­eras
• Record video stream and syn­chro­nized meta­da­ta

pdf

Advanced media player

PDrAW is an advanced video view­er for ANAFI Ai media. The view­er sup­ports both stream­ing (RTP/RTSP) and record­ed (MP4) video on Lin­ux, macOS, Android and iOS plat­forms. PDrAW comes as a library (libp­draw), a wrap­per library (libp­draw-back­end), and a stand­alone exe­cutable (pdraw).

PDrAW also man­ages video meta­da­ta. On the ANAFI Ai, both streamed and record­ed video include meta­da­ta that is pub­licly avail­able and doc­u­ment­ed, allow­ing advanced aer­i­al video pro­cess­ing.

CAD Models

Par­rot pro­vides 3D mod­els of its drones to be inte­grat­ed into your CAD design for rapid pro­to­typ­ing and acces­so­ry inte­gra­tion.

MAVLink and GUTMA Compatibility

ANAFI Ai is com­pat­i­ble with the stan­dard open source MAVLink v1 pro­to­col, which allows real-time data exchange between the UAV and the con­trol sta­tion. The ANAFI Ai can be flown man­u­al­ly or using auto­mat­ic flight plan­ning from a MAVLink com­pat­i­ble base sta­tion such as QGround­Con­trol.

*QGround­Con­trol Inter­face

ANAFI Ai is com­pat­i­ble with the stan­dard open source GUTMA pro­to­col for flight data. The Glob­al UTM Asso­ci­a­tion is a con­sor­tium of major play­ers in the air traf­fic man­age­ment indus­try.

The largest partner ecosystem for Parrot drones

Par­rot con­tin­ues to expand the glob­al ecosys­tem of drone soft­ware providers through its SDK Part­ner Pro­gram, offer­ing tai­lored and unique solu­tions to the grow­ing and evolv­ing needs of pro­fes­sion­al users.

Parrot for every business need

To get the most out of the inspec­tion process, Par­rot offers its cus­tomers an ecosys­tem of com­pat­i­ble drone soft­ware, from fleet man­age­ment, flight plan­ning and flight log import, to sur­vey­ing, map­ping, mon­i­tor­ing and analy­sis of acquired data.

PIX4D × Parrot

Appli­ca­tions for 2D/3D mod­el­ing. A unique set of mobile, desk­top and cloud appli­ca­tions for pho­togram­me­try.

Verizon x Skyward x Parrot

4G LTE enabled drone solu­tion in the Unit­ed States. The first of its kind ANAFI Ai robot­ic drone con­nect­ed to the Ver­i­zon 4G LTE net­work with Sky­ward soft­ware pre-installed. Opens the way for near real-time data trans­fer, remote deploy­ment and non-line-of-sight flights.

Skyward × Parrot

Inte­gra­tion and learn­ing with­in the ANAFI poten­tial. Avi­a­tion man­age­ment plat­form includ­ing air­space, access to LAANC, and train­ing, equip­ment and con­nec­tiv­i­ty to deploy an enter­prise drone fleet.

DroneSense × Parrot

Build, man­age and scale your drone pro­grams. Pub­lic safe­ty drone pro­fes­sion­als can take advan­tage of Drone­Sense’s full suite of capa­bil­i­ties, specif­i­cal­ly designed for the needs of first respon­ders.

Partners in the Parrot ecosystem

Parrot Affiliate Program

The devel­op­er offers to expand the capa­bil­i­ties of its drone pro­gram by inte­grat­ing its SDK.

cyber security

Data Privacy

The devel­op­er notes that it does not col­lect any data with­out the con­sent of users. The deci­sion whether or not to trans­fer data to the Par­rot infra­struc­ture rests sole­ly with the users them­selves. The data host­ed by Par­rot allows the user to syn­chro­nize flight data and flight plans between dif­fer­ent devices, and facil­i­tates sup­port and enables Par­rot to improve its prod­ucts.

ANAFI Ai com­plies with the Euro­pean Union Gen­er­al Data Pro­tec­tion Reg­u­la­tion (GDPR) and goes fur­ther, for exam­ple, it allows you to delete all data in 1 click, thus giv­ing users the eas­i­est con­trol pos­si­ble. This one-click issue is resolved in the FreeFlight7 mobile app or in the pri­va­cy set­tings of your Parrot.Cloud account. Thus, users at any time can not only stop exchang­ing data, but also eas­i­ly request their dele­tion.

If the user con­sents to the exchange of data, their pro­cess­ing will be car­ried out in a com­plete­ly trans­par­ent man­ner, as detailed in the Par­rot Pri­va­cy Pol­i­cy.

When ANAFI Ai is con­nect­ed to the Sky­con­troller 4 via 4G, the Par­rot infra­struc­ture is used to pair the drone and remote con­troller. If the user is not logged into the Parrot.Cloud account, he can still use the 4G con­nec­tion with a unique tem­po­rary account. When using the Par­rot infra­struc­ture for 4G pair­ing, the video is encrypt­ed using a key agreed between the air­craft and the remote con­troller, Par­rot does not have access to unen­crypt­ed videos.

FIPS140‑2 compliant and CC EAL5+ Secure Element Certified

The ANAFI Ai incor­po­rates a Wisekey secu­ri­ty ele­ment that is NIST FIPS140‑2 Lev­el 3 com­pli­ant and Com­mon Cri­te­ria EAL5+ cer­ti­fied. A sim­i­lar safe­ty ele­ment is also built into the Sky­con­troller 4 con­trol equip­ment.

Secu­ri­ty ele­ment func­tions:

• Per­forms cryp­to­graph­ic oper­a­tions
• Stores and pro­tects con­fi­den­tial infor­ma­tion

It uses a per­son­al ECDSA key with domain para­me­ters P521, unique for each drone. It can­not be retrieved from a secu­ri­ty ele­ment. And the cer­tifi­cate asso­ci­at­ed with this key is signed by a cer­tifi­cate author­i­ty.

Wisekey pro­tects the integri­ty of the firmware, pro­vides unique drone iden­ti­fi­ca­tion for 4G pair­ing and strong authen­ti­ca­tion, and unique­ly dig­i­tal­ly signs pho­tos tak­en by the drone.

Secure 4G connection and strong authentication

When the user turns on the 4G con­nec­tion, the stan­dard Wi-Fi con­nec­tion is pri­mar­i­ly used, which is used for the process of secure pair­ing of devices. Dur­ing this process, the user secure­ly con­firms that he is con­nect­ed to a par­tic­u­lar drone. Thanks to the ANAFI Ai Secure Ele­ment, it can do this with­out enter­ing a pass­word inside the drone.

Next, the Par­rot servers reg­is­ter the con­nec­tion between the user and the drone. When the Wi-Fi con­nec­tion between the user and the drone is lost, ANAFI Ai will auto­mat­i­cal­ly switch to 4G con­nec­tion. ANAFI Ai per­forms strong authen­ti­ca­tion against Par­rot servers using its pri­vate key stored on the Secure Ele­ment. Par­rot servers look for relat­ed users and pro­vide pair­ing between ANAFI Ai and Sky­con­troller 4.

To pro­tect the drone con­trol and video streams com­ing to the Sky­con­troller 4, ANAFI Ai sup­ports TLS, DTLS and SRTP pro­to­cols.

Secure provisioning and updating

The sequence of the drone boot process is secure: the sys­tem ver­i­fies that it is using the Par­rot soft­ware and that the soft­ware has not been tam­pered with. A secu­ri­ty check is per­formed at each ini­tial­iza­tion. The Update Ser­vice also con­trols the dig­i­tal sig­na­ture of soft­ware updates.

Setting up custom keys on the Secure Element

ANAFI Ai users have access to a spe­cial Secure Ele­ment oper­a­tor account. This account is used to con­fig­ure keys that are rel­e­vant for this user. Users can con­fig­ure the pub­lic keys of flight providers they trust in the Secure Ele­ment. ANAFI Ai will only per­form flight mis­sions that are signed with these keys. This process pre­vents attack­ers from per­form­ing mali­cious flight mis­sions on the drone.

Photos digitally signed

ANAFI Ai’s Secure Ele­ment can dig­i­tal­ly sign pic­tures tak­en by a drone. This sig­na­ture is proof that:

• Said signed image was tak­en by said unmanned aer­i­al vehi­cle.
• Nei­ther the image itself nor its meta­da­ta has been processed (whether vol­un­tar­i­ly or not) — meta­da­ta, also known as EXIF ​​and XMP, con­tains infor­ma­tion about the date, time, and loca­tion of the image.

In oth­er words, a dig­i­tal sig­na­ture pro­tects all data relat­ed to an image, includ­ing the loca­tion and time it was tak­en and by which ANAFI Ai drone.

Users, as well as part­ners offer­ing soft­ware solu­tions using drone pho­tos, can ver­i­fy the dig­i­tal sig­na­ture of ANAFI Ai pho­tos either with the drone cer­tifi­cate or through a pub­lic key cat­a­log pro­vid­ed by Par­rot.

Transparency and continuous security checks with Bug bounty

When­ev­er pos­si­ble, Par­rot uses stan­dard pro­to­cols and file for­mats. There is no obfus­cat­ed code, no hid­den fea­tures. This allows the user to under­stand how Par­rot prod­ucts work and to test their safe­ty. In addi­tion, Open­Flight, the soft­ware used to con­trol the drone, is open source, so users have full con­trol.

In April 2021, Par­rot launched the “Bug Boun­ty” pro­gram with YesWe­Hack, Europe’s first crowd­sourced secu­ri­ty plat­form. Through this part­ner­ship, Par­rot is lever­ag­ing YesWe­Hack­’s exten­sive cyber­se­cu­ri­ty research com­mu­ni­ty to iden­ti­fy poten­tial vul­ner­a­bil­i­ties in its drones, mobile apps and web ser­vices.

The Bug Bounty program takes place in two stages:

The pri­vate pro­grams ini­tial­ly pro­vide exclu­sive access to select secu­ri­ty researchers and include future Par­rot drone mod­els. The expe­ri­ence and diverse skills of the researchers will con­firm the high lev­el of secu­ri­ty of the prod­ucts before they go to mar­ket, which will help improve the secu­ri­ty of Par­rot users and pro­tect their data.

After this first phase of the pri­vate Bug Boun­ty pro­gram and once com­mer­cial­ized, the prod­ucts move into the pub­lic pro­gram. Their secu­ri­ty is then rig­or­ous­ly reviewed by the entire YesWe­Hack com­mu­ni­ty of over 22,000 cyber­se­cu­ri­ty researchers.

Data management

The Par­rot Cloud allows users who choose to share their data to man­age flight and fleet data, as well as mul­ti­me­dia data received by their drones.

Collected data

The Par­rot cloud col­lects 4 types of data:

“Static” (product data):

• Drone ser­i­al num­ber
• Bat­tery ser­i­al num­ber
• Firmware Ver­sion for Drone and Bat­tery
• Air­craft hard­ware ver­sion and bat­tery
• Device Mod­el
• FreeFlight 7 release ver­sion

“Developments”

• Alerts: bat­tery, autopi­lot, sen­sors
• Con­nec­tion: connect/disconnect, start stream­ing, inter­fer­ence warn­ing, weak sig­nal
• Cam­era: stream­ing sta­tis­tics, set­tings changes
• Flight: state change (take­off, land­ing, hov­er­ing, etc.), flight task acti­va­tion (flight plan, pho­togram­me­try)

“Context Images”

• Time­lapse snap­shots (once every 2 min­utes)
• Deep learn­ing (match­ing more objects, ter­rain for improved flight auton­o­my, track­ing, avoid­ing obsta­cles)
• Stereo vision (depth map)
• Event trig­gered images
• Begin­ning and end of pre­cise hov­er­ing for­mu­la­tions, pre­cise land­ing
• drone crash
• When trans­fer­ring faces are auto­mat­i­cal­ly blurred

“Telemetry”

End use of collected data

Par­rot col­lects and uses data only from cus­tomers who have agreed to share it in order to improve the qual­i­ty of their prod­ucts.

Maintenance management

Pre­ven­tive main­te­nance: our tools col­lect all infor­ma­tion relat­ed to mis­sions (mis­sion type, take­off and land­ing times, num­ber of mis­sions, drone loca­tion, flight speed, flight plan and AirS­DK set­tings). This allows you to receive accu­rate infor­ma­tion about the sta­tus of the ANAFI Ai drone fleet (as well as their con­trollers and bat­ter­ies) in real time.

Cor­rec­tive main­te­nance: The col­lect­ed infor­ma­tion is use­ful for quick­ly deter­min­ing the sta­tus of a par­tic­u­lar drone or bat­tery.

Improving artificial intelligence (AI)

The arti­fi­cial intel­li­gence ele­ments that ANAFI Ai is equipped with (PeleeNet, con­vo­lu­tion­al net­works, etc.) offer users unsur­passed ser­vices and fea­tures: obsta­cle avoid­ance, tar­get fol­low­ing, mul­ti­ple flight modes. The qual­i­ty of AI depends on the quan­ti­ty and qual­i­ty of the col­lect­ed data (images and videos): this data feeds machine learn­ing. In this regard, data qual­i­ty is not the only deci­sive ele­ment: the meta­da­ta asso­ci­at­ed with these data is also of fun­da­men­tal impor­tance. For this rea­son, our tool col­lects images and meta­da­ta on a reg­u­lar basis and depend­ing on events, for a total of 30 to 50 MB per minute.

Flight characteristics

Aerodynamic characteristics

Drives and aerodynamics

CFD image of ANAFI Ai pro­peller blades

The new bio­mimicry pro­peller blades have a lead­ing edge that resem­bles the shape of hump­back whale fins. This approach made it pos­si­ble to increase the propul­sion effi­cien­cy of the pow­er plant, which at the out­put pro­vid­ed an increase in thrust at the same rota­tion speed. This effect is com­pa­ra­ble to an increase in the rotor diam­e­ter.

Acoustic noise is also reduced, espe­cial­ly tonal noise com­ing from the lead­ing edge. There­fore ANAFI Ai is qui­eter [71.5 дБ SPL (A) на высоте 1 м]than Sky­dio 2 [76.4 дБ SPL (A) на высоте 1 м].

• Flight time over 32 min­utes
• Thanks to the opti­mized aero­dy­nam­ic char­ac­ter­is­tics of the fuse­lage and the ANAFI Ai pow­er plant, the max­i­mum speed in for­ward flight is 61 km/h, in side flight and rear­ward flight 58 km/h.
• Max. wind resis­tance 12.7 m/s.
• Thanks to the high torque effi­cien­cy of the motor/propeller and the high auton­o­my of the bat­tery, the range is 22.5 km (at a con­stant speed of 50 km/h in calm weath­er).

CFD images of ANAFI Ai beams

Sensors

To ensure a safe flight, ANAFI Ai is equipped with:

• 2 × IMUs (ICM-40609‑D and ICM42605)
• Mag­ne­tome­ter LIS2MDL
• GPS mod­ule UBX-M8030
• TI OPT3101 time-of-flight (ToF)
• Barom­e­ter LPS22HB
• Ver­ti­cal Cam­era

Sensor Specifications

Flight IMU: ICM-40609‑D

• 3 axis gyro­scope
• Range: ± 2000°/s
• Res­o­lu­tion: 16.4 LSB/°/s
• Offset/accuracy: ± 0.05°/s (after ther­mal and dynam­ic cal­i­bra­tion)
• 3‑axis accelerom­e­ter
• Range: ±16g
• Res­o­lu­tion: 2.048 LSB/mg
• Offset/accuracy: ± 0.5 mg (XY) ± 1 mg (Z) (after ther­mal and dynam­ic cal­i­bra­tion)
• Tem­per­a­ture con­trol: con­trolled heat­ing sys­tem in rela­tion to the ambi­ent tem­per­a­ture, sta­bi­lized with­in: ± 0.15°C
• Mea­sure­ment fre­quen­cy: 2 kHz

Magnetometer: LIS2MDL

• Range: ±49.152G
• Res­o­lu­tion: 1.5mG
• Offset/accuracy: ± 15 mG (after com­pen­sa­tion, at max­i­mum motor speed)
• Mea­sure­ment fre­quen­cy: 100 Hz

Barometer: LPS22HB 1

• Range: 260 to 1260 hPa
• Res­o­lu­tion: 0.0002 hPa
• Offset/accuracy: ± 0.1 hPa
• Tem­per­a­ture con­trol: con­trolled heat­ing sys­tem in rela­tion to the ambi­ent tem­per­a­ture, sta­bi­lized with­in: ± 0.2°C
• Mea­sure­ment fre­quen­cy: 75 Hz
• Mea­sure­ment noise: 20cm RMS

GNSS: UBX-M8030 1

• 25 x 25 x 4mm ceram­ic patch anten­na with +2dB gain improve­ment over ANAFI 1
• Sen­si­tiv­i­ty: cold start ‑148 dBm / track­ing and nav­i­ga­tion: ‑167 dBm
• Time-To-First-Fix: 40 sec­onds
• Accuracy/Accuracy: Posi­tion (stan­dard devi­a­tion 1.4 m), Veloc­i­ty (stan­dard devi­a­tion 0.5 m/s)

Vertical Camera

• Matrix for­mat: 1/6 inch
• Res­o­lu­tion: 640 × 480 pix­els
• Glob­al shut­ter sen­sor
• black and white
• FOV: Hor­i­zon­tal FOV: 53.7°/Vertical FOV: 41.5°
• Focal length: 2.8 mm
• Mea­sure­ment of the speed of move­ment of the opti­cal flow on the ground at 60 Hz
• Point of inter­est cal­cu­la­tion for fine hov­er­ing at 15 Hz and fine land­ing at 5 Hz

ToF: TI OPT3101

• Range: 0–15m
• Res­o­lu­tion: 0.3mm
• Accu­ra­cy: ± 2 cm (after cal­i­bra­tion)
• Mea­sure­ment fre­quen­cy: 64 Hz

Vertical Camera IMU: ICM-42605

• 3 axis gyro­scope
• Range: ±2000°/s
• Res­o­lu­tion: 16.4 LSB/°/s
• Offset/accuracy: ± 0.1 °/s (after dynam­ic cal­i­bra­tion)
• 3‑axis accelerom­e­ter
• Range: ±16g
• Res­o­lu­tion: 2,048 LSB/mg
• Offset/accuracy: ± 2.0 mg (XY) ± 5.0 mg (Z) — after dynam­ic cal­i­bra­tion
• Mea­sure­ment fre­quen­cy: 1 kHz
• Hard­ware syn­chro­niza­tion with ver­ti­cal cam­era, accu­ra­cy: 1 µs

Autopilot

The ANAFI Ai flight con­troller pro­vides sim­ple and intu­itive pilot­ing: no train­ing is required to oper­ate it. It allows you to auto­mate many flight modes (Flight Plan, Cam­era­man, Hand take-off, Smart RTH). Sen­sor fusion algo­rithms com­bine data from all sen­sors to esti­mate the atti­tude, alti­tude, posi­tion, and veloc­i­ty of ANAFI Ai.

Con­di­tion assess­ment is essen­tial for the prop­er func­tion­ing of drones. Quad­copters are inher­ent­ly unsta­ble when the flight con­troller is used in open loop; to eas­i­ly con­trol them, not to men­tion autonomous con­trol, it is nec­es­sary to sta­bi­lize them using closed-loop con­trol algo­rithms. These algo­rithms cal­cu­late and send the com­mands need­ed by the ANAFI Ai to the engines to achieve the desired tra­jec­to­ries.

Indoor flight

In the absence of a GPS sig­nal, ANAFI Ai relies heav­i­ly on ver­ti­cal cam­era mea­sure­ments to esti­mate speed and posi­tion. The ver­ti­cal cam­era per­for­mance is deter­mined by two main algo­rithms:

• Opti­cal flow for speed esti­ma­tion
• Detec­tion and match­ing of key points for posi­tion esti­ma­tion

The ver­ti­cal cam­era algo­rithms can work in low light con­di­tions due to the fact that ANAFI Ai is equipped with a pair of LED lights locat­ed next to the ver­ti­cal cam­era. They allow the drone to main­tain sta­bil­i­ty, espe­cial­ly when fly­ing indoors or in con­di­tions with­out GPS, at a height of less than 5 m above the ground. The pow­er of the LED lights adapts auto­mat­i­cal­ly, depend­ing on the needs of the algo­rithm.

autonomous flight

Key Features

• Swivel­ing sens­ing sys­tem with wide field of view
• Selec­tion of the depth of the sur­round­ing space based on stereo match­ing and depth by motion
• Rep­re­sent­ing the Envi­ron­ment as an Infill Grid
• Autonomous detec­tion and avoid­ance of obsta­cles at speeds up to 29 km/h

This chap­ter details the sen­sors, hard­ware, and algo­rithms used by the ANAFI Ai to enable autonomous flight. It is orga­nized as fol­lows:

1. Detailed descrip­tion of the ANAFI Ai per­cep­tion sys­tem
2. Per­cep­tion algo­rithms used to recon­struct the 3D envi­ron­ment sur­round­ing drones
3. Replan­ning and avoid­ing obsta­cles

Perceptual system strategy

The per­cep­tion of the 3D envi­ron­ment is a key abil­i­ty to achieve autonomous flight, espe­cial­ly in con­fined spaces. It is a con­di­tion for guar­an­teed detec­tion and avoid­ance of obsta­cles, which reduces the load on the drone oper­a­tor, increas­es the suc­cess of the mis­sion and ensures the safe­ty of the air­craft.

Unleash­ing the full poten­tial of a fly­ing cam­era that can move freely and rotate in all direc­tions with­out restric­tion requires an effi­cient sens­ing solu­tion. In par­tic­u­lar, the per­cep­tion sys­tem must be able to receive infor­ma­tion about the envi­ron­ment in direc­tions that cor­re­spond to trans­la­tion­al move­ment in flight — regard­less of the ori­en­ta­tion of the cam­era.

ANAFI Ai relies on a unique tech­ni­cal solu­tion based on two mechan­i­cal sus­pen­sions to sep­a­rate the ori­en­ta­tion of the main cam­era and the per­cep­tion sys­tem:

• The main cam­era is mount­ed on a 3‑axis gim­bal, which makes its 3D ori­en­ta­tion inde­pen­dent of the drone’s ori­en­ta­tion.
• The per­cep­tion sys­tem is mount­ed on a sin­gle-axis gim­bal — con­nect­ed to the drone’s yaw move­ment, it can be ori­ent­ed in any direc­tion.

Dual ANAFI Ai sta­bi­liz­er for per­cep­tion and visu­al­iza­tion.

The tilt axes of the two gim­bals are collinear and com­bined to achieve an ultra-com­pact design.

Thanks to this solu­tion, it is pos­si­ble to direct the main cam­era and the per­cep­tion sys­tem in two dif­fer­ent direc­tions. This design avoids the use of cost­ly cam­eras on the sides, top, bot­tom and rear of the drone while pro­vid­ing a large avail­able field of view for the sens­ing sys­tem.

This sec­tion is orga­nized as fol­lows:

1. Detailed infor­ma­tion about the sen­sors used for the sens­ing sys­tem
2. Spec­i­fi­ca­tions for both the main cam­era gim­bal and the sens­ing sys­tem gim­bal
3. Per­cep­tion Sys­tem Ori­en­ta­tion Strate­gies to Har­ness the Poten­tial of the Dual Sus­pen­sion Design

Sensors

The per­cep­tion sys­tem is based on a pair of iden­ti­cal cam­eras with the same pitch axis.

1‑axis mechan­i­cal joint of ANAFI Ai per­cep­tion sys­tem.

The specifications of the sensors are as follows:

• Mod­el: Onse­mi AR0144CSSM28SUD20
• Col­or: mono­chrome
• Res­o­lu­tion: 1280×800 pix­els
• Frame rate: 30fps
• glob­al shut­ter
• Total hor­i­zon­tal field of view: 118° (110° used for per­cep­tion)
• Total ver­ti­cal field of view: 72° (62° used for per­cep­tion)
• Focal length: 1.47 mm (0.039 inch­es — 492.94610 pix­els)
• Aper­ture: f/2.7

The technical characteristics of the stereo pair are as follows:

• Com­mon Pitch Axis
• Baseline/Distance: 62 mm (2.44 inch­es)
• Syn­chro­nous cap­ture at 30fps

double suspension

The main cam­er­a’s mechan­i­cal gim­bal is a 3‑axis pitch-roll-yaw mechan­i­cal hinge with the fol­low­ing fea­tures:

• Pitch Lim­iters: ‑116°/+176°
• Roll lim­iters: +/- 36°
• Yaw Lim­iters: +/- 48°

The sens­ing sys­tem mechan­i­cal gim­bal is a sin­gle-axis gim­bal with the fol­low­ing char­ac­ter­is­tics:

• Pitch lim­iters: ‑107°/+204°
• Trav­el time from one end stop to anoth­er: 300 ms

The per­cep­tion sys­tem has 311° trav­el (of which 296° is not masked by the body of the drone), which allows for rear per­cep­tion as well.

Instan­ta­neous ver­ti­cal field of view and lim­iters of the ANAFI Ai per­cep­tion sys­tem.

The system has been designed in such a way that:

• Pro­peller blades do not fall into the field of view of the main cam­era
• The main cam­era does not cov­er the field of view of the per­cep­tion sys­tem
• Main cam­era and sens­ing sys­tem can be ful­ly fold­ed back to pro­tect lens­es, dur­ing stor­age or in case of an in-flight emer­gency

Hor­i­zon­tal field of view of the ANAFI Ai per­cep­tion sys­tem.

When lean­ing back, the per­cep­tion sys­tem moves to the high­est posi­tion, there­by pro­vid­ing a clear view.

Per­cep­tion of ANAFI Ai in ful­ly reclined posi­tion for back­ward flight.

Environmental reconstruction

Recon­struc­tion of the 3D envi­ron­ment for autonomous flight is per­formed in two stages:

1. Extract­ing depth infor­ma­tion from per­cep­tion as depth maps
2. Com­bin­ing depth map data into a 3D infill grid

Two meth­ods are used to obtain depth maps from per­cep­tion sen­sors:

1. Depth based on stereo match­ing
2. Depth from motion

Depth from Stereo Match

The main method for extract­ing depth infor­ma­tion is based on par­al­lax between the two stereo cam­eras of the per­cep­tion sys­tem. By pho­tograph­ing the envi­ron­ment in the same direc­tion but from two dif­fer­ent posi­tions, objects in the field of view of the per­cep­tu­al sys­tem appear in dif­fer­ent posi­tions in the images tak­en by the two cam­eras. The clos­er the object, the greater the dif­fer­ence in posi­tion.

Thus, the strat­e­gy is to deter­mine the points in the images tak­en by the left and right stereo cam­eras cor­re­spond­ing to the same object in the field of view of the per­cep­tu­al sys­tem, and mea­sure the dif­fer­ence in the posi­tion of these points in the two images. This dif­fer­ence is called dis­par­i­ty and is mea­sured in terms of pix­els.

An illus­tra­tion of the prin­ci­ple of stereo vision — a red 3D dot is in dif­fer­ent posi­tions on the left and right images.

The dis­par­i­ty can then be relat­ed to the depth of each of these points using the fol­low­ing rela­tion­ship: depth = focal (focal length) * base­line (base­line) / dis­par­i­ty (dis­par­i­ty), where the depth and base­line are expressed in the same units, and the focal length and dis­par­i­ty are expressed in num­ber of pix­els.

The result of the cal­cu­la­tion takes the form of a 176 × 90 pix­el depth map, for which the val­ue of each pix­el cor­re­sponds to a depth in meters. The depth map is updat­ed at a rate of 30 Hz.

An exam­ple of an image tak­en with the right cam­era of the ANAFI Ai sens­ing sys­tem (left) and the cor­re­spond­ing depth obtained using stereo map­ping (right). The col­ormap tran­si­tions from red (Clos­er) to pur­ple (Far­ther) — white means “Out of reach”.

The imme­di­ate result of this is that the depth mea­sured by this method is dis­cretized, since dis­par­i­ty can only take on dis­crete val­ues ​​(num­ber of pix­els). A 3D point that is suf­fi­cient­ly far from the per­cep­tu­al sys­tem to pro­duce a the­o­ret­i­cal dis­par­i­ty of less than one pix­el would be con­sid­ered infi­nite, since the cor­re­spond­ing actu­al, dis­crete dis­par­i­ty would be 0. The accu­ra­cy of the stereo match­ing method decreas­es with increas­ing dis­tance, although there are meth­ods to reduce this phe­nom­e­non by achieve sub­pix­el sam­pling.

In addi­tion, the dis­par­i­ty diverges as depth approach­es zero. Since the num­ber of pix­els in images is lim­it­ed, the dis­par­i­ty val­ue is also lim­it­ed. As a con­se­quence, there is a min­i­mum depth at which the per­cep­tu­al sys­tem is blind. This min­i­mum depth val­ue is 36 cm for ANAFI Ai.

About cal­i­bra­tion: each pair of stereo cam­eras is fac­to­ry cal­i­brat­ed to accu­rate­ly mea­sure small dis­crep­an­cies that may exist between two cam­eras and com­pen­sate for them in the onboard depth cal­cu­la­tion.

The user can also re-cal­i­brate a pair of stereo cam­eras using the test pat­tern sup­plied with the drone. In par­tic­u­lar, to some extent, the drone is able to detect poten­tial cal­i­bra­tion errors that may occur dur­ing its life­time. In this case, the air­craft soft­ware will attempt to adjust and com­pen­sate for them, and if it fails, a noti­fi­ca­tion will appear ask­ing you to recal­i­brate.

Depth from motion

Drone move­ment can also be used to col­lect images of the envi­ron­ment from dif­fer­ent angles and thus recov­er depth infor­ma­tion. This method is called motion depth or monoc­u­lar per­cep­tion because a sin­gle mov­ing cam­era is enough to col­lect depth infor­ma­tion.

The prin­ci­ple of oper­a­tion is sim­i­lar to stereo vision, but instead of com­par­ing images of the envi­ron­ment obtained by dif­fer­ent observers at the same time, per­cep­tion com­pares images of the envi­ron­ment obtained by the same observ­er at dif­fer­ent times. If the drone is mov­ing, then images from this unique observ­er will be tak­en from dif­fer­ent angles. Know­ing the posi­tion at which each frame was tak­en, it is pos­si­ble to tri­an­gu­late points cor­re­spond­ing to the same fea­ture in dif­fer­ent images and return them to 3D.

The out­put is a three-dimen­sion­al point cloud con­tain­ing up to 500 points for ANAFI Ai, gen­er­at­ed at a fre­quen­cy of 10 Hz.

An exam­ple of a point cloud gen­er­at­ed from motion depth — the col­ormap tran­si­tions from red (Near­er) to pur­ple (Far­ther).

The motion-based depth algo­rithm in ANAFI Ai typ­i­cal­ly gen­er­ates less infor­ma­tion (sparse point cloud) than the stereo match­ing algo­rithm and requires the air­craft to move to col­lect infor­ma­tion. In addi­tion, this algo­rithm can­not extract infor­ma­tion in the exact direc­tion of motion (at least for live broad­casts), because in this direc­tion the objects in the images look almost sta­tion­ary (expan­sion focus).

How­ev­er, it also has a bet­ter detec­tion range (the­o­ret­i­cal­ly infi­nite range) than the stereo match­ing method.

Infill grid

Depth infor­ma­tion derived from the stereo and monoc­u­lar per­cep­tion algo­rithms is inte­grat­ed into the fill grid. This grid dis­cretizes the 3D envi­ron­ment into 3D cubes called vox­els. Each vox­el is assigned a prob­a­bil­i­ty of being occu­pied by an obsta­cle or, con­verse­ly, being free of an obsta­cle.

A ray trans­mis­sion algo­rithm is used to inte­grate depth infor­ma­tion into the infill mesh. For each pix­el of the depth map obtained as a result of stereo match­ing, it is con­vert­ed to a three-dimen­sion­al point, and for each point of the point cloud — to the depth obtained as a result of move­ment:

• In the fill grid, a ray is drawn from the posi­tion of the per­cep­tion sys­tem to the posi­tion of the 3D point.
• The prob­a­bil­i­ty of occu­pan­cy of a vox­el con­tain­ing a 3D point increas­es.
• The prob­a­bil­i­ty of occu­pan­cy of all vox­els crossed by the ray, except for the one con­tain­ing the 3D point, decreas­es.

In this way, the ret­i­cle acts as a tem­po­ral fil­ter for depth infor­ma­tion, absorb­ing any poten­tial depth mea­sure­ment noise, and as a mem­o­ry of pre­vi­ous mea­sure­ments, allow­ing nav­i­ga­tion in com­plex envi­ron­ments even in the absence of a con­tin­u­ous 360° field of view of the per­cep­tion sys­tem.

Fill grid exam­ple. Over­laid on the right stereo cam­era view are vox­els with a high degree of con­fi­dence that they are occu­pied, rang­ing in col­or from red (Near) to pur­ple (Far).

The infill grid is the basis for the motion plan­ning algo­rithms used by ANAFI Ai for autonomous flight and obsta­cle avoid­ance.

obstacle avoidance

With the knowl­edge of the 3D envi­ron­ment of the drone’s envi­ron­ment stored in the infill grid, it is pos­si­ble to pro­vide the ANAFI Ai with the abil­i­ty to avoid obsta­cles. This pro­vides sig­nif­i­cant addi­tion­al safe­ty for autonomous mis­sions, but is also use­ful for man­u­al flight, espe­cial­ly if the line of sight between the pilot and the drone is degrad­ed.

Every 30 ms, ANAFI Ai pre­dicts what the nom­i­nal flight path will be on a short time hori­zon in the future. This pre­dic­tion is derived from links sent by the user, whether it be pilot com­mands from the con­sole, way­points to join a flight plan, or an input tra­jec­to­ry. Then, using the sim­u­lat­ed drone inter­nal mod­el, the replan­ning algo­rithm cal­cu­lates the small­est pos­si­ble cor­rec­tions to this pre­dict­ed nom­i­nal tra­jec­to­ry that make it col­li­sion-free and fea­si­ble for the drone.

An exam­ple of a cor­rect­ed tra­jec­to­ry com­put­ed by the obsta­cle avoid­ance algo­rithm in response to a ref­er­ence tra­jec­to­ry col­lid­ing with a tree.

The ANAFI Ai obsta­cle avoid­ance sys­tem has been designed to oper­ate at speeds up to:

• Lev­el flight: 29 km/h
• Climb: 14 km/h
• Descent: 11 km/h

Avoid­ance per­for­mance is lim­it­ed in rain or strong wind con­di­tions, in low light con­di­tions, or in con­di­tions of impaired satel­lite nav­i­ga­tion. In addi­tion, before fly­ing, you must make sure that the lens­es of the per­cep­tion sys­tem are clean.

Flight missions

Key Features

The Air SDK (see SDK sec­tion) allows devel­op­ers to access every drone sen­sor, cam­era, con­nec­tion inter­face, and stand­alone func­tion. There­fore, they can cus­tomize the behav­ior of the drone to cre­ate flight mis­sions. Each flight mis­sion con­tains a set of basic behav­iors or modes:

• On the ground: behav­ior when the motors are stopped, e.g. sen­sor cal­i­bra­tion.
• Take­off: var­i­ous take­off strate­gies
• Hangup: fixed point hold
• Flight: man­u­al and autonomous flight func­tions
• Land­ing: var­i­ous land­ing strate­gies
• Crit­i­cal: when a crit­i­cal con­di­tion is detect­ed

Cus­tom flight mis­sions can cre­ate new behav­iors or reuse them from the default mis­sion.

Price and availability

Accord­ing to the lat­est offi­cial infor­ma­tion, ANAFI Ai will be avail­able for pur­chase from Jan­u­ary 2022. The cost of the drone will be deter­mined by the equip­ment. The stan­dard kit will cost the con­sumer $5,476 (tax includ­ed).

Documentation for download

Doc­u­men­ta­tion for famil­iar­iza­tion with the prod­uct from the devel­op­er:

1. Down­load full tech­ni­cal doc­u­men­ta­tion from the devel­op­er
2. Down­load the prod­uct spec­i­fi­ca­tion from the devel­op­er

Video

Top ANAFI Ai Reviews from Developer and Users

Unpack­ing and first flight.

Test­ing 4G con­nec­tiv­i­ty on Par­rot Anafi Ai.

Demon­stra­tion of the unique obsta­cle avoid­ance sys­tem based on mobile stereo cam­eras from Par­rot.

Drone take­off and land­ing test from a mov­ing vehi­cle.

Installing a 4G SIM card.

Test flight of a drone at a dis­tance with a 4G con­nec­tion. The total flight time was 29 min­utes at an aver­age hor­i­zon­tal flight speed of 58 km/h (16 m/s). Dur­ing this time, the drone was able to over­come 27 km. You can also notice that the flight was car­ried out in strong wind con­di­tions, which the drone sys­tem con­stant­ly warned about.

3D house mod­el­ing in 1 click.

An exam­ple of a drone in 3D map­ping mode.

5 km of pow­er lines are mapped in 3D.

3D map­ping of the Ital­ian city of Posi­tano.

To real­ize this pharaon­ic project, Par­rot drew on the expe­ri­ence of Erwan Renaudin of the French com­pa­ny Logiroad, an expert in the cre­ation of dig­i­tal copies of cities. An ANAFI Ai drone has mapped 100 acres of an urban canyon. In total, the drone took 4,800 48-megapix­el pho­tos. All flights were car­ried out in 4G mode to avoid inter­fer­ence in the city. To cre­ate the final 3D mod­el, Erwan had to use three dif­fer­ent soft­ware: Pix4Dmapper, Metashape and Real­i­ty Cap­ture. The final 3D mesh con­sists of 800 mil­lion tri­an­gles.

The flight of the ANAFI Ai to the top of the 368-meter active TV tow­er Berlin­er Fernse­hturm, dur­ing which the con­fi­dent oper­a­tion of the machine is demon­strat­ed in con­di­tions of heavy inter­fer­ence.

4G pilot­ing in a metrop­o­lis.

360-degree pho­to by ANAFI Ai.

Auto­mat­ic flight and shoot­ing along a planned route.

Demon­stra­tion of auto­mat­ic take­off and land­ing on a mov­ing ship.