A Multiple Hypothesis Tracking Approach to Collision Detection for Unmanned Aerial Vehicles

It is obvious that future unmanned aerial vehicle (UAV) applications will have a high degree of automation, up to autonomy, and will be integrated into civil airspace to go beyond line of sight. To ensure the safe operation of these systems, new technologies are required. Currently, the regulatory and technological frameworks are insufficient. For instance, there are no standards for sensors required to detect obstacles and avoid possible collisions. It is evident that data from several sensors need to be combined to obtain a robust environmental map in order to cope with the high standards of aviation. This paper aims to introduce a Multiple Hypothesis Tracking (MHT) approach to multi-sensor data fusion for future collision avoidance systems. The data from several sensors are merged to obtain new and more accurate measurement data to capture the environment more robustly. The sensors consist of several optical and thermal sensors, radar and transponder systems. This combined measurement data is processed by an avoidance algorithm to calculate avoidance maneuvers. So far, simulations and ground based tests have shown that the implemented MHT approach provides qualitatively better results than conventional probabilistic data fusion approaches. In a next step, test flights will be performed to evaluate the proposed MHT data fusion approach in an airborne environment.