DYNAMIC OBSTACLE CHARACTERIZATION AND AVOIDANCE FOR UNMANNED AERIAL SYSTEMS

Abstract

To address the challenge of avoiding dynamic obstacles during the course of the 2018 SUAS competition, a multistage obstacle characterization and avoidance algorithm was designed and implemented. The obstacle characterization section begins with simple base assumptions about behavior and goes through several more advanced stages of obstacle characterization and prediction as more data arises and advanced behavior is detected. The path finding section of the algorithm uses a recursive Monte Carlo path sampling function with a flexible structure that allows for usage with varying computational budgets. It also restricts its computational usage depending on the level of variability in the obstacles.