Distance-Based Heuristic Solvers for Cooperative Path Planning with Heterogeneous Agents

Abstract

Cooperation among different vehicles is a promising concept for transportation services. For instance, vehicle platooning on highways with self-driving vehicles is known to decrease fuel consumption. This makes us construct paths on graphs, where many vehicles share sub-paths to form platoons. The platooning model has been recently generalized to model different types (e.g., trucks and UAVs) but existing exact solvers using Integer Programming (IP) are not experimentally evaluated on various settings and heuristic solvers are underdeveloped; hence solving large instances remains difficult. In this study, we propose heuristic solvers using a distance-based grouping of vehicles. Our solver first finds groups of vehicles and constructs paths in each group independently. We also experimentally investigate both exact and heuristic solvers. Experimental results suggest that IP-based solvers only solve small instances due to the overhead of compiling IP models. We observed that our solvers were almost fifth magnitude faster than the exact solver with at most 25% additional travel costs. Also, our method achieved roughly 15% additional costs if requests are clustered in terms of their locations, meaning that distance-based heuristic solvers could find moderate solutions within typically a few seconds.