Scholarlink

To mitigate position sensor failures in multi-electric aircraft systems operating under extreme conditions such as high temperature, high pressure, and turbulent vibrations, this paper presents a sensorless position control strategy based on a Model Reference Adaptive System (MRAS) enhanced with deep reinforcement learning. Specifically designed for Permanent Magnet Synchronous Motors (PMSM), the strategy integrates a Twin Delayed Deep Deterministic Policy Gradient (TD3) controller within an Actor-Critic framework to improve dynamic performance and reduce output angle tracking errors caused by MRAS estimation inaccuracies and suboptimal control parameters. The controller’s real-time online learning capability and extensive data-driven training confer enhanced robustness against various system disturbances, ensuring overall system stability. Simulation experiments validate the effectiveness and reliability of the proposed control strategy, demonstrating its potential for application in demanding aerospace environments.

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

TD3 Deep Reinforcement Learning-Based Improved Sensorless MRAS Control Strategy for Multi-Electric Aircraft PMSM

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Han Wu,
  • Bingqiang Li,
  • Xiong Zhou,
  • Dongheng Wang,
  • Yuening Deng

摘要

To mitigate position sensor failures in multi-electric aircraft systems operating under extreme conditions such as high temperature, high pressure, and turbulent vibrations, this paper presents a sensorless position control strategy based on a Model Reference Adaptive System (MRAS) enhanced with deep reinforcement learning. Specifically designed for Permanent Magnet Synchronous Motors (PMSM), the strategy integrates a Twin Delayed Deep Deterministic Policy Gradient (TD3) controller within an Actor-Critic framework to improve dynamic performance and reduce output angle tracking errors caused by MRAS estimation inaccuracies and suboptimal control parameters. The controller’s real-time online learning capability and extensive data-driven training confer enhanced robustness against various system disturbances, ensuring overall system stability. Simulation experiments validate the effectiveness and reliability of the proposed control strategy, demonstrating its potential for application in demanding aerospace environments.