Improvements of the ARTIS Sampling-based Path Planner for Obstacle Field Navigation in Urban Terrain

This work studies an online roadmap-based path planner for an unmanned rotorcraft that is enhanced with a sensorbased trajectory time reduction concept. A recent evaluation of showed that linear roadmap paths under-perform with respect to the enroute trajectory time in urban terrain scenarios. The challenge addressed in this paper is the tradeoff between a runtime efficient but coarse linear online roadmap and computationally more demanding nonlinear collision free path smoothing. An online connection strategy is required that considers motion constraints and the obstacle sensor's field of view, in order to still allow for feasible connections to the roadmap at higher velocities and in narrow corridors. Our approach proposes a finite horizon connection strategy that considers special cases during hover as opposed to sole forward flight. During the enroute flight, a clamped piecewise cubic spline interpolation is computed in real time and closely fitted along linear path corridors. Simulation results indicate that the proposed approach reduces the trajectory time while the computational overhead is kept low. It benefits from the roadmap's coarse free space representation such that the computational overhead is minimized. The approach considers approximate rotorcraft kinematics and the obstacle sensor's field of view for increased motion safety. The concept is compared to benchmark baseline performances where a significant trajectory time reduction is achieved.