Collision Avoidance for Micro Air Vehicles

The innovative use of variational methods to dynamically segment scenes, which leads to a fast, natural approach to estimating the location of unknown 3-D obstacles is proposed. The previously developed algorithm has been shown in simulation to be suitable for real-time processing in flight using current generation processors, and to be robust in the presence of transient sensor data, distortion, and obscuration. It is employed to rapidly construct a 3-D database in the flight path ahead of the vehicle, and is combined with custom-developed guidance laws to produce a real-time in-flight obstacle detection and avoidance capability using only a sequence of 2-D images. This technology is to be tailored to application on the Block II production Wasp micro air vehicle and evaluated in near-real-time simulation. Simulation results are to be validated using flight video collected on the Wasp. Sensitivity studies will be conducted to develop a set of design requirements, and a preliminary design for hardware and software implementation completed. Detailed design, development, flight test on the Wasp, algorithm refinement, and demonstration at an MOUT test site will be carried out in phase II.