Exploiting innovative sensor data fusion techniques for Sense and Avoid units to be installed on-board Unmanned Aerial Systems

Sense and Avoid Systems play an important role on-board Unmanned Aerial Vehicles in order to be allowed flying into civil Airspace. The key idea of these systems is to detect obstacles in the own trajectories, tracking the detected objects and execute a collision avoidance manoeuvre if the obstacle is closely approaching, thus becoming a collision threat. These functions can be achieved defining an adequate sensor setup, choosing a dynamic model that allows describing the target motion properly, identifying a suitable filtering methodologies given the non-linearities in the dynamic model. This thesis deals with identification and test of innovative sensor data fusion techniques to be implemented in a fully autonomous system devoted to avoidance of non-cooperative intruders. In particular, sensors, hardware and software architectures are described, focusing the attention on the impact of an innovative filtering methodology, such as Particle Filter, on the performance of the developed tracking software with respect to assessed technique, such as Extended Kalman Filter. The Particle Filter Obstacle Detect and Tracking system has been developed and tested in off-line simulations based on real data gathered during a flight test campaign within TECVOL project (carried out in collaboration with the Italian Aerospace Research Center). In order to evaluate the effectiveness of the developed software for the assessment of a collision risk, an analysis has been carried out for the estimation of the Distance at Closest Point of Approach. Numerical results have shown that the Particle Filter algorithm is able to provide performance comparable to the Extended Kalman Filter ones and allows obtaining some improvements with respect to the EKF in terms of DCPA, thus reducing the delay in the collision detection