无人艇航行规则及运动学约束下的改进RRT轨迹规划方法

现有的RRT算法没有考虑无人艇的运动特性,难以解决无人艇轨迹规划问题,也没有基于无人艇航行规则来考虑无人艇的动态避碰。针对上述问题,在无人艇航行规则及运动学约束下,提出了改进的双层RRT动态轨迹规划方法。在第一层框架中,改进了探索点及步长选择策略,并结合国际海上避碰规则公约与最短会遇时间建立最优位置窗口来构造四向扩展随机树,从而可以在考虑海事规则的前提下快速搜索出联通路径。在第二层框架中,考虑到无人艇的运动学约束,将上一层的联通路径点作为分段启发点,然后结合速度运动模型来限制无人艇的拐角与转弯半径,并基于速度运动模型得到的弧长计算出每一个节点的代价值,最终在动态障碍物环境中得到一条可行平滑轨迹。仿真与实船实验均验证了该改进算法的有效性,实验表明该改进算法可以有效地解决传统RRT算法离障碍物过近、路径不平滑、不符合无人艇运动学与无人艇航行规则等问题。其中,轨迹转折数目为0,与障碍物最近距离是传统RRT算法的两倍以上,最大转折角度指标远好于传统RRT算法。

Improved RRT Trajectory Planning Method under Unmanned Surface Vehicle Rules and Kinematic Constraints

The RRT algorithm which does not consider the motion characteristics of the unmanned surface vehicle(USV) and dynamic collision avoidance based on unmanned navigation rules, is hard to solve the trajectory planning problem of USV. In view of the above problems, an improved dynamic trajectory planning method of double-layer RRT under the navigation rules and kinematic constraints of USV is proposed in this paper. In the first layer framework, the exploration point and step selection strategy is improved. In combination with the convention on the international regulations for preventing collision at sea(COLREGS) and the time to closest point of approaching(TCPA), the optimal position window is established and a four-way extended random tree is constructed. Thus, the connection path can be quickly searched out on the premise of considering maritime rules. In the second tier framework, considering the kinematic constraints of the USV, the connecting path point of the upper layer is taken as the segment heuristic point, and then the corner and turning radius of the USV are limited by combining the velocity motion model. What''s more, based on the arc length obtained from the velocity motion model, the cost value of each node is calculated, and finally a feasible smooth trajectory is obtained in a dynamic obstacle environment. Simulation and actual ship tests verify the effectiveness of the proposed algorithm. The experiment shows that the improved algorithm can effectively solve the problems, such as the traditional RRT algorithm is too close to obstacles, the path is not smooth, and does not conform to the kinematics and navigation rules of USV. What''s more, the number of turns is zero, the closest distance to obstacles is more than twice that of traditional RRT and the turning angle index are far better than it.