Abstract
This paper introduces the Cooperative Electric Vehicles Planning Problem (CEVPP), which consists in finding a path for each vehicle of a fleet of electric vehicles, such that the global plan execution time (including travel time, charging time and waiting time) is minimal (e.g., by limiting the number of vehicles who need to charge simultaneously at the same charging station, which leads to waiting time). We show that the strategy which consists in planning each possible permutation of EVs and keeping the one providing the best solution is not only time intractable, but also not optimal. We propose different centralized planning algorithms to solve CEVPP instances: (1) a baseline non-cooperative CEVPP planner, (2) an optimal cooperative planner that finds a solution inside a carefully designed state space, and (3) multiple variants of an approximate cooperative planner based on the Cooperative-A* algorithm. We compare the solutions’ quality and computation times obtained by these CEVPP planners. Our empirical results show that our best approximate cooperative EV planner found solutions with a reasonably small computational overhead compared to the baseline algorithm. The solutions found by our cooperative planners had significantly lower plan execution time globally, including travel time, waiting time and charging time, than the solution found by our baseline non-cooperative planner. On average, our empirical results show that our cooperative algorithms decreased the global (including each EVs) waiting time by more than 90%, while having a negligible impact on the charging and driving time.
Jaël Champagne Gareau
PhD Student in Computer Science
My research interests include AI, data structures, algorithms and HPC.