Time to impact estimations as a measure for relaxed decision making in multi-UAV deconfliction

This work presents a decentralized scheme for conflict detection and resolution among Unmanned Aerial Systems (UAS). An estimate of the time-to-collision is provided for different possible headings of an agent. This time-to-collision is used as a measure for decision to avoid certain maneuvers which based on the performance of the agents can lead to loss of separation. In our work we consider non-uniform, non-holonomic and cooperative planar agents scenario. According to the original formulation of the Velocity Obstacle method, for any agent facing one or more opponent agents, we can define a region for the velocity vector that does not lead to a conflict. However, the simple application of this method does not provide any information on the likelihood of the conflict occurring whenever we choose a non free from collision direction. With the time-to-collision (TC) method based evaluation of the danger, we developed a relaxed decision technique that may allow an agent to safely perform an originally forbidden maneuver. Numerical simulations of this scheme have been conducted that prove the effectiveness of the proposed method and the performance enhancements resulting from it.