The aim of this project is to build a navigation framework for an unmanned aerial vehicle, and to apply the result to the unmanned helicopter within the BEAR project. Navigation schemes that rely on a perfect knowledge of the environment are naturally unreliable. We consider this assumption unrealistic. We will divide the navigation problem into two tightly coupled subproblems: sensing and planning.
Sensing the environment is an essential requisite for any navigation problem. We'll make use of computer vision and gps/ins sensors and combine them in order to acquire a confident knowledge of the environment and of our motion in it. Sensor redundancy will also assess robustness with respect to fault detection and handling.
Motion planning, i.e., the capability of deciding what motion to execute in order to reach a goal, is fundamental to the design and realization of autonomous robots. A good motion planning algorithm must ensure safety, by avoiding collision and planning feasible paths, and efficiency (minimum travel time, etc.).