Reliable Global Path Planning of Off-Road Autonomous Ground Vehicles Under Uncertain Terrain Conditions
Reliable Global Path Planning of Off-Road Autonomous Ground Vehicles Under Uncertain Terrain Conditions
复制标题
不确定地形条件下越野自主地面车辆可靠全局路径规划
DOI：
10.1109/tiv.2023.3317833
复制
发表时间：
2024-01
影响因子：
8.2
通讯作者：
Jianhua Yin;Lingxi Li;Z. Mourelatos;Yixuan Liu;D. Gorsich;Amandeep Singh;Seth Tau;Zhen Hu
中科院分区：
文献类型：
--
作者：
Jianhua Yin;Lingxi Li;Z. Mourelatos;Yixuan Liu;D. Gorsich;Amandeep Singh;Seth Tau;Zhen Hu
路径规划在确保越野自主地面车辆（AGV）高效、安全运行方面起着至关重要的作用。目前的方法主要集中在最小化的行程时间或路径长度的AGV，往往忽略了这样一个事实，即AGV可能会失败，在许多方面在操作过程中，由于随机和粗糙的地形条件。本文的目的不仅是为车辆生成合适的路径，而且确保规划的路径对于复杂地形条件的车辆来说总体可靠、安全。为了实现这一目标，本文开发了一种基于可靠性的使命规划方法越野AGV受到两种故障模式的移动性（最大可达速度和车辆垂直加速度）引起的不确定的地面属性的地形。首先建立了基于物理的车辆动力学仿真模型，预测了给定道路地形条件下的车辆机动性，然后考虑越野地形条件下的不确定性因素，采用代理建模方法对AGV的机动性可靠性进行了分析。然后，将两种失效模式的可靠性约束与快速探索随机树星星（RRT*）算法相结合，确定一条最优路径，即满足两种失效模式可靠性要求的最短路径。最后通过一个实例验证了所提方法在考虑地形不确定性的情况下进行路径规划的有效性。
Path planning plays a vital role in ensuring the efficient and safe operation of off-road autonomous ground vehicles (AGVs). Current methods mostly focus on minimizing the travel time or path length of AGVs and often overlook the fact that the AGVs could fail in many ways during the operation due to stochastic and rough terrain conditions. The objective of this article is to not only generate a proper path for the vehicle but also ensure that the planned path is overall reliable and safe for the vehicle with complex terrain conditions. To achieve this goal, this article develops a reliability-based mission planning method for off-road AGVs subject to two failure modes in terms of mobility (the maximum attainable speed and vehicle vertical acceleration) induced by the uncertain ground properties of the terrain. A physics-based vehicle dynamics simulation model is first employed to predict vehicle mobility for any given terrain conditions of a path. Mobility reliability of an AGV is then analyzed using surrogate modeling methods considering uncertainty sources in the off-road terrain conditions. After that, the reliability constraints for the two failure modes are integrated with the Rapidly-exploring Random Tree Star (RRT*) algorithm to identify an optimal path, which is the shortest path while satisfying the reliability requirements of the two considered failure modes. Results of a case study demonstrated the effectiveness of the proposed methods for path planning with the consideration of uncertainty in the deformable terrain.