To assist in aerial collision avoidance of non-cooperative threats, we have developed an innovative approach to predict aircraft trajectories given current state estimates from an airborne radar. The approach is based on quantifying state estimation uncertainty and maneuver prediction uncertainty. The composite uncertainty can fully characterize the state-dependent covariance around the current estimate through sampling in the state space. This paper describes a threat aircraft state prediction method building upon MIT-Lincoln Lab's Uncorrelated Encounter Models and forward-propagated Probabilistic Reachable Sets (PRS) that describe regions of differing collision risk. The PRS represent a joint probabilistic characterization of the collision risk regions, and are used along with Maximum Reachable Sets (MRS) to define the worst-case regions. Under ongoing research, we apply the knowledge of reachability to estimate total risk of collision or near mid-air collision with the threat vehicle over a time horizon. We have integrated the above measures and tested within medium-fidelity flight simulations.
展开▼
