Mixed-Integer Energy Management for Multi-Motor Electric Vehicles with Clutch On-Off: Finding Global Optimum Efficiently

This article introduces a novel approach to energy management in multi-motor electric vehicles, leveraging mixedinteger model predictive control (MI-MPC). First, an energy management strategy is proposed to co-optimize torque allocation and decoupling decisions, minimizing both energy consumption and frequency of clutch engagement changes. Secondly, to address computational challenge inherent in solving the resultant mixed-integer (MI) problem, a bi-level programming approach is proposed. In this approach, the torque allocation subproblem is efficiently solved at the inner level with explicit closed-form analytical solution, while the outer level optimizes clutch decisions through implicit dynamic programming (i-DP). Evaluation in a high fidelity virtual environment shows energy savings exceeding 4% compared to heuristic controllers prevalent in modern electric vehicles. The i-DP based solution process guarantees finding global optimum for the MI problem in every MPC update. The presented strategy shows an average solution time of 1 ms in a laptop, conceptually indicating its real-time potential and possible integration in multi-motor electric vehicles.