Title: Fast 3D Collision Avoidance for Autonomous Vehicles based on Trigger Time Selection (Instructor: Prof. Xu, Huan)
UAS is a fast-growing industry with extensive economic implications and will eventually be integrated into the national airspace system (NAS), which will require UAS to have the capability of quick collision avoiding with other air vehicles. This proposal and the following thesis research contains following parts:
First, presents an eﬃcient 3D collision avoidance algorithm named Fast Geometric Avoidance algorithm (FGA), for ﬁxed wing Unmanned Aerial Systems (UAS). The algorithm, increases the ability of aircraft operations to complete mission goals by enabling fast collision avoidance of multiple obstacles: shortens the UAS collision avoidance duration by selection of the proper avoidance starting time tc. Where tc is computed based pm the UAS kinematic, conﬂicts likelihood map, and navigation constraints. This operation enables the update path to be as closer as possible to the UAVs resume designed path, decreasing the length of path variation and the respond time cost. In comparison to a current geometry method and the waypoint generation method, the algorithm shows 40% to 90% of reduction in computational time for the same obstacle avoidance scenarios.
Secondly, research on the successful rate and the constraints of the geometry collision avoidance algorithm. The sensor detection distance; the relative velocity ratio; the altitude; the heading, pitching angle (solid angle for 3D situation) of the UAV and obstacles will impact the exist of the active collision avoidance solutions. Monte Carlo simulation and statistic method is used here to validate the exactly percentage of the above factors to the successful rate contribution. Thirdly, with the time-varying conﬂicts possibility map P(x,y,z,t) of the UAV surrounding environment and the diﬀerent reward function J(w,a,b), where w, a, b present the roll angle, heading and pitching angle of the UAV mechanism, to compute a more applicable avoidance starting time tc under complicated scenarios to get the most safety and the least cost path. Monte Carlo simulations, ﬂight missions in an aircraft simulator and the hardware ﬁxed-wing aircraft experiments validate the algorithm eﬀectiveness with successful results.