A probabilistic framework for unmanned aircraft systems collision detection and risk estimation

The airborne collision detection is a challenging problem due to inherent noise in onboard sensor(s), limited computational resources available for unmanned aircraft system and variability in intruder dynamics resulting in uncertainty in predicting intruder future trajectories. In this paper, we develop an innovative approach to quantify likely intruder trajectories and estimate the probability of collision risk for a pair of aircraft flying at the same altitude in close proximity. The proposed approach is formulated in a probabilistic framework building upon the uncorrelated encounter model (UEM) developed by MIT Lincoln Laboratory (LL) and the concept of forward reachable sets. The performance of proposed approach is evaluated using Monte Carlo based simulations where statistically relevant encounter scenarios are sampled from the MIT LL UEM.