Physical self-awareness for a machine
Over the next few years, University of Delft researcher Coen de Visser will be working on a system that would give autonomous robots and drones a degree of “physical self-awareness.” The idea is to make machines aware of their physical capabilities even in the event of a failure. This should lead to the fact that in the future, autonomously working robots and drones will be able to independently get out of an emergency.
4SENCE
The project revolves around the further development of a new algorithm called 4SENCE. This algorithm is a mathematical representation of physical self-consciousness in the “Safe frame” format. As soon as the drone begins to threaten unpredictable behavior as a result of a failure, 4SENCE is triggered to safely land the drone or complete the mission as safely as possible.
De Visser: “A robot equipped with 4SENCE is effectively aware of its physical limitations and thus can make intelligent decisions about how to deal with small and large failures. Take, for example, an inspection drone that lost one of its motors. This invariably results in an uncontrolled fall with potentially fatal consequences. With 4SENCE, such a catastrophic situation could have been avoided and the drone could have made a controlled forced landing or, if possible, within the safety framework, even completed the initial inspection with less performance.”
Tipping points
The study is unique because De Visser establishes a connection with another area of research, namely the area of Critical transitions (Critical transitions), also known as Tipping points (Tipping points). “Until now, comparable algorithms have been used to a limited extent due to the large uncertainty of the models used. But with the help of Tipping points, these uncertainties can be minimized. This makes it easier to predict the “safe structure,” says De Visser.
To support his research project in drone technology, De Visser received an €800,000 VIDI grant from the Netherlands Research Organization (NWO). The study will be carried out at the Faculty of Aerospace Engineering at Delft University of Technology in collaboration with researchers from the University of Oxford and Wageningen University.