Scholarlink

<p>The trajectory prediction layer of autonomous vehicles predicts the further trajectory of adjacent vehicles and plans paths to avoid potential risks based on ego-vehicle and environmental perception data. However, traditional path planning algorithms based on simplified traffic models and static predictions fail to capture the nonlinear and time-varying characteristics of actual traffic flows, making it difficult to adapt to dynamic traffic rules and complex traffic environments, leading to inaccurate planning results. To address these issues, a multi-model switching law integrated with Viterbi decoding is proposed to obtain the driving intentions of adjacent vehicles and predict their trajectories. An RRT* four-step optimization algorithm is designed to plan the path of ego-vehicle in complex environments. The Prescan-Carsim-Simulink results demonstrate that the ego-vehicle can avoid obstacles based on the results of multi-cycle path planning in static obstacle avoidance scenarios. The vehicle executes a lane change 2&#xa0;m in advance in leading vehicle cut-in scenarios. The vehicle experiences a 34.4% reduction in the peak of yaw rate during overtaking, with heading changes becoming smoother and the extreme values being reduced by 19.3%. The curvature extreme point is 90.04, with an extreme value of − 1.07&#xa0;m in multi-vehicle interaction scenarios. The vehicle completes lane-changing actions earlier and reduces the yaw displacement. The strategies proposed in this paper can respond promptly and optimize lane-changing actions, enhancing the safety and predictability of path planning.</p>

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

Research on path planning of intelligent vehicle integrated with adjacent vehicle prediction

  • 期刊论文

  • 发表日期：

    2025年9月22日

查看全文
  • Xiang Fu,
  • Qianfeng Ruan,
  • Maojia Tang,
  • Jiaqi Wan,
  • Zitai Xiao,
  • Yipeng Yin,
  • Xilong Zhang,
  • Tianqi Yang

摘要

The trajectory prediction layer of autonomous vehicles predicts the further trajectory of adjacent vehicles and plans paths to avoid potential risks based on ego-vehicle and environmental perception data. However, traditional path planning algorithms based on simplified traffic models and static predictions fail to capture the nonlinear and time-varying characteristics of actual traffic flows, making it difficult to adapt to dynamic traffic rules and complex traffic environments, leading to inaccurate planning results. To address these issues, a multi-model switching law integrated with Viterbi decoding is proposed to obtain the driving intentions of adjacent vehicles and predict their trajectories. An RRT* four-step optimization algorithm is designed to plan the path of ego-vehicle in complex environments. The Prescan-Carsim-Simulink results demonstrate that the ego-vehicle can avoid obstacles based on the results of multi-cycle path planning in static obstacle avoidance scenarios. The vehicle executes a lane change 2 m in advance in leading vehicle cut-in scenarios. The vehicle experiences a 34.4% reduction in the peak of yaw rate during overtaking, with heading changes becoming smoother and the extreme values being reduced by 19.3%. The curvature extreme point is 90.04, with an extreme value of − 1.07 m in multi-vehicle interaction scenarios. The vehicle completes lane-changing actions earlier and reduces the yaw displacement. The strategies proposed in this paper can respond promptly and optimize lane-changing actions, enhancing the safety and predictability of path planning.