Track A*: Fast Visibility-Aware Trajectory Planning for Active Target Tracking

Track A* (TA*) is a new offline trajectory planner from researchers Hanxuan Chen, Kangli Wang, and Ji Pei, designed for active target tracking. It operates on a discretized 4D spatio-temporal grid (x, y, z, t), combining a layered Directed Acyclic Graph search with three engineering optimizations: cross-time obstacle distance caching against a Bounding Volume Hierarchy (BVH), per-layer beam pruning, and a configurable multi-ray visibility evaluator. This allows TA* to efficiently find high-quality tracking trajectories while trading strict theoretical optimality for practical scalability.

In a 1000-scenario stress test across eight CARLA maps, TA* converged on all scenarios in just 45 seconds using 32 workers. Compared to an unoptimized priority-queue A* baseline, TA* reduced mean planning time by 23.0x and worst-case planning time by 11.8x, while raising convergence from 56.9% to 100%. On the subset of 141 baseline-converged scenarios, average visibility changed by only -0.15 percentage points, with no scenario exceeding a 5 pp drop. The paper positions TA* as a practical offline reference planner for building multi-modal tracking datasets and benchmarking online planners, though it notes limitations in dense vegetation environments like Town07.