K-Means Partitioned Space Path Planning (KPSPP) for Autonomous Robotic Harvesting

Abstract

A three-dimensional coverage path-planning algorithm is proposed for discrete harvesting machines. Although prior research has developed methods for coverage planning in continuous-crop fields, no such algorithm has been developed for discrete crops such as trees. The problem is formulated as a graph traversal problem and solved using graph techniques. Paths to facilitate autonomous operation are generated. A case study is formed around the novel tree-to-tree felling system developed by the University of Canterbury and Scion. This machine is being developed to manoeuvre through New Zealand's plantation forest to fell Pinus radiata trees on steep ( ≤ 45°) terrain. Algorithm performance is evaluated in 14 commercial plantation forests. Results indicate that a mean coverage of 84.43% was achieved.

A Comprehensive Review of Coverage Path Planning in Robotics Using Classical and Heuristic Algorithms

“…The local search and the cost path could be improved by utilizing the nearest neighbor insertion algorithm and variable neighborhood strategy. Meaclem et al [314] and Ding et al [315] used the k-means clustering method and the density-based spatial clustering algorithm, respectively to partition the regions and assign the robots in each region for area coverage. Azpurua et al [32] segmented the environment into sub-hexagonal cells and divided them into sub-regions by the k-means algorithm.…”