Aerial Chasing of a Dynamic Target in Complex Environments

QP Chaser: Polynomial Trajectory Generation for Autonomous Aerial Tracking

“…To the best knowledge of the authors, there are few studies considering target visibility as vital importance in dynamic environments. The approach in [20] makes a bunch of polynomial motion primitives and select the best paths that satisfy safety and visibility constraints, and it can be applied to dynamic environments. However, as target occlusion is imminent, the planner of [20] may alter the chasing path to a region with path topology, where inevitable occlusion occurred.…”

“…The approach in [20] makes a bunch of polynomial motion primitives and select the best paths that satisfy safety and visibility constraints, and it can be applied to dynamic environments. However, as target occlusion is imminent, the planner of [20] may alter the chasing path to a region with path topology, where inevitable occlusion occurred. Also, [21] designed visibility cost to address occlusion by dynamic obstacles.…”

“…ôc * 0 (t) := c * 0 (t) − ô(t). As in (20), we approximate a nonpolynomial term ôc * 0 (t) 2 to a polynomial d3 (t). With the approximated the term d3 (t), the collision constraint is defined as follows:…”

“…Since trigonometric terms in (31), (33) are non-polynomial, the Lagrange interpolation is used as (20). The approximated c * (t) is denoted as c * (t).…”

“…We perform AirSim [33] simulation for a benchmark test of our method and [20], [24]. In real-world experiments, stereo cameras ZED2 [34] are mounted on the drone, and intel NUC and Jetson Xavier NX are equipped as onboard computers.…”