The growing adoption of connected and automated vehicles has led to significant advancements in intelligent transportation. The computational efficiency stands as a critical bottleneck in the commercialization of connected and automated vehicles. This paper proposes a novel transfer learning-enhanced multi-objective predictive control strategy to mitigate the adverse effects of computational burden for vehicular platoon. A data-driven platoon model is established with subspace identification to characterize the non-ideal driving behavior and complex powertrain structure of electric vehicles. To balance multi-objective conflicts among driving safety, driving comfort and energy economy, a multi-objective cost function incorporating the predictive sequence is designed. Then, a grey wolf optimizer is devised to guide the search process, with the goal of achieving globally optimal trade-offs. Here, the knowledge of vehicle in the source domain, such as data-driven model and controller hyperparameters, is transferred to vehicles in the target domain. Based on the knowledge, each vehicle in the target domain just should update the knowledge in terms of the individual dynamic characteristic and tasks. With in this mind, a transfer learning-enhanced multi-objective predictive control strategy is developed to enhance the performance of tasks in the target domain. Finally, a hardware-in-the-loop experiment platform with the Carmaker and the driving simulator is conducted. The experimental results demonstrate that our proposed transfer learning-enhanced multi-objective predictive control strategy could improve the 28.8% computational efficiency while ensuring the satisfied platoon tracking performance.
Wu et al. (Thu,) studied this question.