Recurrent Neural Network Based Multi-Robot Route Planning for Steep-Land Harvesting Systems

Terrains with steep inclines cannot be utilized for crop production as it is hazardous to operate conventional farm equipment for various cropping tasks. This research, which is a part of a larger effort to deploy a team of robots for seeding and other similar tasks, proposes an approach for route planning of a team of ground mobile robots (drones) for seed delivery as well as an unmanned aerial vehicle (UAV) to continually replenish the drones with seeds. The overall cost is formulated in terms of linearly constrained integer quadratic programming (LCICQ). Using a Lagrangian function, a recurrent neural network with primal and dual sets of neurons is proposed to converge to optimal solutions.