基于高精地图的障碍物轨迹预测与误差评估

Obstacle Trajectory Prediction and Error Evaluation Based on High Definition Map

为确保在复杂的交通环境下自动驾驶车辆行驶的安全性和舒适性,不仅需要感知周围障碍物当前时刻的行驶状态,更需要对其未来的行驶状态进行预测。高精地图(High Definitionmap,HD)的应用可以为自动驾驶系统提供丰富的道路信息。提出了一种基于高精地图的障碍物轨迹预测算法,较为准确地预测了障碍物未来的行驶轨迹,使得车辆可以根据障碍物未来7 s的轨迹而做出相应的决策调整,提升了自动驾驶安全性及舒适性。提出了一种多维度的轨迹精度评估方法,从多种维度评估了预测精度,反映了轨迹预测算法在不同方面的表现。

To ensure the safety and comfort of automateddriving vehicles in complex traffic environments, it is not only necessary to perceive the vehicle running state of surrounding obstacles at the current moment, but also predict the vehicle running state of surrounding obstacles in the time ahead. The application of High Definition map(HD) provides rich road information for the automated driving system. Therefore, this paper proposed an obstacle trajectory prediction algorithm based on HD map, which accurately predicts the trajectory of obstacles in the time ahead, so that the vehicle can make corresponding decision adjustment according to the trajectory of obstacles in the next 7 s, and improve the safety and comfort of automated driving. A multi-dimensional trajectory accuracy evaluation method is also proposed. The prediction accuracy is evaluated from various dimensions, which reflects the performance of trajectory prediction algorithm in different aspects.