A heuristic Dual sampling domain Reduction-based Optimal Rapidly-exploring Random Tree scheme is proposed by guiding the planning procedure of the optimal rapidly-exploring random tree (RRT*) method through learning environmental knowledge. The scheme aims to plan low fuel expenditure, easy-execution, and low collision probability paths online for an unmanned surface vehicle (USV) under constraints. First, an elliptic sampling domain, which is subject to an elliptic equation and the shortest obstacle avoidance path estimation, is created to plan short paths. Second, by the consideration of the USV motion states, obstacles and external interferences of the current, the near sampling domains of tree nodes are reduced to exclude high-cost sampling domains. Path feasibility is ensured by explicitly handling motion constraints. Third, a safe distance-based collision detection (CD) scheme and a velocity-based bounding box of USV are proposed to decrease the path collision probability. Additionally, a layered USV online path planning framework is built in accordance with the model predictive control method, and the path smoothing scheme is applied via the Dubins curve under the curvature constraint. Results demonstrate that the proposed dual sampling domain reduction method outperforms traditional reduction schemes in terms of improving the execution efficiency of RRT*. Meanwhile, the proposed CD method is more reliable than the conventional one.

通过学习环境知识，指导最优快速探索随机树（RRT *）方法的规划过程，提出了一种基于启发式双采样域约简的最优快速探索随机树方案。该计划旨在为约束条件下的无人水面车辆（USV）在线规划低油耗，易于执行和低碰撞概率路径。首先，创建一个椭圆采样域，该域受一个椭圆方程和最短避障路径估计的影响，以规划短路径。其次，考虑到USV运动状态，电流的障碍和外部干扰，减少了树节点的近采样域，以排除高成本的采样域。通过明确处理运动约束来确保路径的可行性。第三，为了降低路径碰撞概率，提出了一种基于距离的安全碰撞检测（CD）方案和基于速度的USV包围盒。此外，根据模型预测控制方法构建了分层的USV在线路径规划框架，并在曲率约束下通过Dubins曲线应用了路径平滑方案。结果表明，在提高RRT *的执行效率方面，所提出的双重采样域缩减方法优于传统的缩减方法。同时，所提出的CD方法比传统方法更可靠。根据模型预测控制方法建立了分层的USV在线路径规划框架，并在曲率约束下通过Dubins曲线应用了路径平滑方案。结果表明，在提高RRT *的执行效率方面，所提出的双重采样域缩减方法优于传统的缩减方法。同时，所提出的CD方法比传统方法更可靠。根据模型预测控制方法建立了分层的USV在线路径规划框架，并在曲率约束下通过Dubins曲线应用了路径平滑方案。结果表明，在提高RRT *的执行效率方面，所提出的双重采样域缩减方法优于传统的缩减方法。同时，所提出的CD方法比传统方法更可靠。