Concept and preliminary study on an online flight path generation method using random search and dynamics filtering

This paper introduces the concept and computer simulation results of a flight path generation method using random search and dynamics filtering. The objective of research is online flyable trajectory generation for aerial exploration in a complex environment that has enormous obstacles and/or way points. Conventional path generation methods utilize SQP, GA and so on. They attempt to satisfy global optimality, dynamic correctness and obstacle avoidance simultaneously. However, most of the search nodes will fail to fly due to numerous obstacles and be wasted. Therefore, the conventional researches require high computation cost. The proposed method resolves flight trajectory generation into two steps to solve in less times; 1) finding path that avoid obstacles, 2) ensuring dynamical correctness. The first step is a random search using sampling based method. It quickly makes a path that avoids obstacles, but the generated path generally lacks dynamical correctness so that aircraft cannot follow it. The second step is a dynamics filtering. The path generated by the first step is modified to be flyable. The modification is carefully set to minimum in order to conserve obstacle avoidance.