Efficient Navigation of a Cartesian Robot Among Plants in a Garden

This paper presents a reactive navigation method for a Cartesian robot operating in dynamic garden environments with growing plants. The key challenge — safe navigation amid continuously changing plant structures — is addressed by extending artificial potential functions (APFs) to 3D workspaces. The environment, including crops and fences, is modeled as elliptic cylinders generated from point cloud data. The APF is reformulated to handle these three-dimensional obstacles, and a heuristic is introduced to reduce local minima. Unlike traditional decoupled motion strategies, our approach enables coordinated 3D motion through simultaneous horizontal and vertical movements. Experiments demonstrate robust, safe navigation as plant canopy and height evolve.