Human-like mechanism deep learning model for longitudinal motion control of autonomous vehicles
Zhenhai Gao,Tong Yu,Fei Gao,Rui Zhao,Tianjun Sun
DOI: https://doi.org/10.1016/j.engappai.2024.108060
IF: 8
2024-07-01
Engineering Applications of Artificial Intelligence
Abstract:Artificial intelligence (AI) plays a critical role in the prediction, planning, and control of autonomous vehicle. The original motion control methods are increasing in accuracy, but their control style markedly differs from that of human drivers. This not only degrades the passenger's ride experience but also brings unfamiliarity to other drivers, leading to safety hazards. Moreover, some of the methods are a direct application of AI techniques without any specific enhancements tailored for vehicle control. This both fails to fully utilize the potential of AI and reduces interpretability. To address these issues, in this paper, deep learning methods are integrated with driver control mechanisms to propose a human-like neural network for vehicle longitudinal dynamics estimation and control (HNN). First, the process of driver estimation and control of vehicle dynamics is proposed and described in mathematical language. Subsequently, the HNN composed of three sub-networks is presented. The three sub-networks correspond to the three sub-processes of the driver's mechanism, which makes the proposed HNN method more applicable to vehicle dynamics control and more interpretable in specific engineering domains. The effectiveness of the HNN method is verified on a real-world dataset. The results demonstrate that the HNN not only enhances vehicle human-like control but also surpasses baselines in terms of control style consistency and convergence rate.
automation & control systems,computer science, artificial intelligence,engineering, electrical & electronic, multidisciplinary