事件驱动的自动驾驶车辆局部避障与跟踪控制

Local Obstacle Avoidance and Tracking Control for Autonomous Vehicles under Resource Constraints

摘要:

针对无人驾驶车辆在跟踪全局路径时遭遇静态障碍物的局部重规划与跟踪控制问题，提出了一种融合事件触发机制的分层模型预测控制（MPC）架构；该架构上层为局部避障规划器，基于车辆点质量模型，通过引入新型避障函数并求解带约束的优化问题，实时生成光滑可行的局部避障路径；下层为路径跟踪控制器，采用线性时变MPC方法，对车辆非线性动力学模型进行实时线性化，并综合多种约束以实现高精度跟踪；为降低计算负荷，创新性地设计了一种基于输出误差的事件触发机制，仅在跟踪偏差超出动态阈值时触发MPC优化求解，有效节省通信资源；最后通过Simulink/Carsim联合仿真在多场景下验证表明，所提方法能可靠规划避障路径，实现稳定、精准跟踪，并有效平衡控制性能与计算效率。

Abstract:

This paper addresses the problem of local re-planning and tracking control for autonomous vehicles encountering static obstacles while following a global path. It proposes a hierarchical model predictive control (MPC) architecture incorporating an event-triggered mechanism. The upper layer of this architecture is a local obstacle avoidance planner, which is based on the vehicle point mass model. By introducing a new obstacle avoidance function and solving an optimization problem with constraints, it generates smooth and feasible local obstacle avoidance paths in real time. The lower layer comprises a path-tracking controller employing a linear time-varying MPC method. This method performs real-time linearisation of the vehicle"s nonlinear dynamic model and integrates multiple constraints to achieve high-precision tracking. To reduce computational load, an innovative event-triggered mechanism based on output error is designed. This mechanism triggers the MPC optimisation solution only when the tracking deviation exceeds a dynamic threshold, thereby significantly enhancing system real-time performance while maintaining control accuracy. Simulink/Carsim co?simulation under multiple scenarios verifies that the proposed method can reliably plan obstacle?avoidance paths, realize accurate and stable tracking, and effectively balance control performance and computational efficiency.