Receding horizon motion planning for multi-rotor with point cloud environment

Abstract

This paper proposes a collision-free receding horizon motion planning for a fixed-altitude maneuvering multi-rotor platform with point cloud obstacle environment, based on model predictive control(MPC). Obstacles surrounding multi-rotor is assumed to given as the 3D point cloud form detected with 3D scanning sensors such as stereo vision sensor or LIDAR(Light Detection and Ranging) sensor. Considering fixed-altitude flight, 2D point cloud information is extracted from the given 3D point cloud information. To reduce the number of constraints on obstacles while maintaining the obstacle structure, 2D point cloud obstacles are clustered into several groups with DBSCAN(Density-Based Spatial Clustering of Applications with Noise) algorithm and polygonized by selecting the verticies to contain all of the points in each clusters. Complete reference path with the obstacle avoidance is generated with Voronoi diagram and Dijkstra algorithm to reach the target from the current position. Receding horizon motion planning is performed based on model predictive control by iteratively solving constrained optimization problems numerically to generate the path and the force commands within the prediction horizon considering the constraints such as dynamics, current states, hardware specifications, reference path, and the surrounding obstacles. As a result of the receding horizon motion planning, sequence of path and force commands considering constraints within the prediction horizon and control horizon are created. With the suggested receding horizon motion planning, one integrated algorithm with sub-algorithms is provided to compute the predicted path and control sequences from the given point cloud environment. Moreover, the computation time is reduced compared with the entire path optimization, while the feasibility to reach the target position is guaranteed through the created safety region based on complete reference path and obstacle information.