Scholarlink

The future of mobility is being built today, autonomous driving and ADAS systems are fundamentally redefining the role of the driver, either by offering increased control or by relieving them of their tasks entirely. By emerging autonomous vehicles, the need for adaptive control becomes essential to ensure safety and performance in dynamic environments. Among control techniques, PID controllers remain a popular solution due to their simple conception and stability of dynamic processes and even the presence of external disturbances. However, their performance is highly dependent on the precise tuning of their parameters. To address the challenge, this paper proposes an innovative approach for lateral control of autonomous vehicles by combining an evolutionary strategy optimization (ESO) PID controller with a radial basis function neural network (RBFNN). The ESO algorithm is used to search for the optimal PID parameters for offline tuning, thus ensuring optimal initial PID gains. Then, an adaptive RBFNN takes over to adjust these parameters according to environmental variations for online adaptation. This architecture allows maintaining precise control even under dynamic conditions. Simulations demonstrate the efficiency of the proposed method in terms of tracking the reference lateral acceleration, compared to PID-ESO and PID MATLAB tuning.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimal Adaptive Neural Network Based PID Using Evolutionary Strategy Optimization for Autonomous Ground Vehicle

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ouakki Bouchiad,
  • Lhoussain El Hajjami,
  • Ahmed Nouaçry,
  • Abdessamad Klilou

摘要

The future of mobility is being built today, autonomous driving and ADAS systems are fundamentally redefining the role of the driver, either by offering increased control or by relieving them of their tasks entirely. By emerging autonomous vehicles, the need for adaptive control becomes essential to ensure safety and performance in dynamic environments. Among control techniques, PID controllers remain a popular solution due to their simple conception and stability of dynamic processes and even the presence of external disturbances. However, their performance is highly dependent on the precise tuning of their parameters. To address the challenge, this paper proposes an innovative approach for lateral control of autonomous vehicles by combining an evolutionary strategy optimization (ESO) PID controller with a radial basis function neural network (RBFNN). The ESO algorithm is used to search for the optimal PID parameters for offline tuning, thus ensuring optimal initial PID gains. Then, an adaptive RBFNN takes over to adjust these parameters according to environmental variations for online adaptation. This architecture allows maintaining precise control even under dynamic conditions. Simulations demonstrate the efficiency of the proposed method in terms of tracking the reference lateral acceleration, compared to PID-ESO and PID MATLAB tuning.