Individualized Clustered Cooperative Communication Units in Automated Electrical Routing in 3D CAD

Many industries are faced with the dilemma of increasing the amount of wires and cables in their products. The process of mapping the path of these cables is tedious, iterative, and prone to errors. Instead of manually specifying all the waypoints for a diverse set of cables, automation can provide globally optimized and proven pathways accelerating the product development process. To implement automated electrical routing, an industry-appropriate method is seamlessly integrated into existing 3D computer-aided design (CAD) workflows. The aim of this research is to evaluate the effectiveness and practicality of multi-agent reinforcement learning in determining the most efficient paths in three-dimensional space. To achieve this goal, information is extracted directly from 3D CAD and the results are immediately fed back into CAD. This paper proposes a novel approach that involves clustering the cables based on example paths prior to the actual learning process. Then, a communicating multi-agent proximal policy optimization (PPO) algorithm learns the routing process. To solve the shortest path problem in three-dimensional space while considering cable-and environment-specific constraints and minimizing the total cable length, the agents' accessible space is restricted to a maximum distance from the initial 3D CAD geometry. The developed approach is explained in this paper and compared to established techniques in electrical routing such as the A* algorithm.