Deep Learning-Based Energy-Efficient Spacing Policy and Platooning Control Co-Design for Connected and Automated Vehicles on Inclined Roads
Hanwen Zhang, Jicheng Chen, Chaojie Zhu, Hui Zhang
Abstract
In the work, we present a novel energy-efficient spacing policy and learning-based distributed model predictive control co-design framework for heterogeneous vehicle platoon considering road slope factor. A nonlinear longitudinal dynamics model with look-ahead slope consideration is established for each vehicle in platoon, whose total energy depletion is computed by a constructed energy consumption model. Benefiting from the co-design of platooning control and spacing policy design, energy efficiency is achieved by optimizing inter-vehicular distances and control inputs simultaneously in a designed energy-efficient distributed model predictive control (EDMPC) problem. To reduce online computational requirement of the proposed EDMPC and realize real-time implementation, deep learning technique is adopted to approximate the implicit controller obtained from EDMPC problem. Finally, simulations are conducted to demonstrate the effectiveness and strength of the proposed algorithm. Comparison studies illustrate over 2% percentage energy reduction and average 2ms execution time can be achieved by the EDMPC framework.