Online Local Cubic Spline Path Smoothing and Trajectory Following

Abstract

The main focus of this work is studying the enhancement of trajectory time for a rotorcraft by generating a local smooth curved path based on the global planner linear path in online real time constrains. The technique used for smoothing is the piecewise cubic spline interpolation. The work identifies the problems presented in having a limited certain volume of geometry defined by the sensor field of view, the need of developing a connection strategy to the global planner, defining the spline conditions to compute a distinct path and post smoothing collision detection and avoidance. A heuristic finite horizon connection strategy is proposed to connect the agent to the global planner with respect to the vehicle field of view. Furthermore, spline initial condition, final condition and path parameterization are defined. Moreover, effective collision detection and avoidance techniques are utilized. Finally, the cubic spline smoothing technique is simulated in urban and simple scenarios. The results are compared against baseline solution for benchmarking. In addition, the results are compared against the linear path following solution. The comparison defines aspects that could enhance the technique and improve the results.