基于非线性模型预测控制的交叉口自动驾驶车辆轨迹规划与控制

Trajectory planning and control of autonomous vehicles at intersections based on nonlinear model predictive control

为保证自动驾驶车辆在交叉口内安全且高效的行驶,本文提出一种基于非线性模型预测控制(NMPC)的交叉口内自动驾驶车辆的轨迹规划与轨迹控制算法。以自动驾驶车辆完成期望的驾驶行为为目标,基于车辆轮廓不互相重叠的原理建立防碰撞约束,并根据交叉口的几何信息构建车道约束;为了保证车辆行驶符合运动学规律,根据交通规则和车辆特性设置运动约束。通过运用公开自然驾驶智能汽车仿真测试环境(Onsite)接口,将该算法在300个由真实世界孪生的测试场景中进行了仿真测试,测试结果表明:本文算分在安全、效率和舒适度3个方面都有良好的表现。

To ensure the safe and efficient operation of autonomous vehicles at intersections,this study proposed a trajectory planning and control algorithm based on Nonlinear Model Predictive Control(NMPC).The algorithm aims to achieve desired driving behavior while considering anti-collision constraints based on non-overlapping vehicle contours and lane constraints derived from geometric information.Additionally,motion constraints are established in accordance with traffic rules and vehicle characteristics to ensure compliance with kinematic laws.The proposed method is evaluated through simulation using 300 real-world twin test scenarios within the publicly available Natural Driving Smart Vehicle Simulation Test Environment(Onsite),demonstrating favorable performance in terms of safety,efficiency,and comfort.