Friday 14 August 2009
The sectors of the defence and security demand more and more successful drones to carry out complex missions.
To answer this request, Probayes collaborates with INRIA to develop software modules which push away the limits of the robustness and the autonomy of decision.
For this purpose ProBT®, the Bayesian inference engine from ProBAYES has been successfully embedded on an ARM processor (Gumstix Overo 600MHz).
The target application is the flight control of a drone micro-helicopter. The CoaX® helicopter commercialized by Skybotix technologies is a 300g machine with 2 coaxial propellers, a video camera, an altimeter, an IMU (Inertial Measurement Unit), 3 proximeters, WIFI and Bluetooth links, a DSP micro-controller and an ARM processor.
INRIA E-motion group has implemented a Bayesian flight control of this helicopter for autonomous take-off, obstacle avoidance, wall falling and landing on a target. We used the robot Bayesian Programming (Bayesian-Programming.org) methodology and tools to do so.
At present, developments are done on learning the behaviors from instances of human controlled flights. This proves that the powerful ProBT® inference engine may run on embedded processors for real-time control tasks.
|