Coverage path planning for agricultural ground multi-robots under complex terrain conditions

Abstract

As steep terrain conditions are considered to be too risky for human operators, agricultural tasks must be carried out by multi-robots. This research considers minimum energy path planning by a team of ground robots for full coverage of uneven terrain. Experimental data from a real robot is used to develop a machine learning model to estimate minimum energy straight line paths between pairs of proximally located points. Longer paths are approximated as sequences of line segments. Exemplar-based clustering is used to identify a set of waypoints, which can be used for sensor placement, to ensure full terrain coverage. Treating the waypoints as the vertices of a digraph, optimal cyclic paths for navigation by a team of agricultural robots are determined. The proposed approach for waypoint identification and optimal path discovery can be implemented through local message passing in a sensor network. Comparison with other recently proposed methods, and simulations using real-world elevation data establish the effectiveness of the proposed approach.