【摘要】 针对机械臂三维路径规划中存在的搜索随机性强、收敛速度慢及计算复杂度高等问题，提出了一种基于空间体素化的动态目标引导边界框的自适应步长双向快速扩展随机树(Spatial Cell-Dynamic Target Guided Bounding box-Adaptive Step-Bidirectional Rapidly-exploring Random Tree star, SC-DTGB-AS-BI-RRT~*)算法。该算法通过多层次优化策略显著提升了路径规划性能。首先，运用空间体素化将三维空间离散化为自由区域和非自由区域，并约束采样于自由区域，从而降低计算复杂度并提高环境感知精度；其次，针对传统双向快速扩展随机树(Bidirectional Rapidly-exploring Random Tree, BI-RRT~*)算法的盲目性，提出动态目标引导机制与动态边界框约束机制协同的优化策略，该策略通过调整迭代次数动态适配目标引导区域，即在算法初期进行全局搜索，中期平衡全局搜索与局部收敛，后期加速路径收敛，同时引入动态边界框约束机制，在障碍物密集区域收缩采样范围，障碍物空旷区域扩展搜索空间，从而减少冗余节点，提高采样效率；再次，提出自适应步长机制，根据环境特征动态调整步长，即在自由区域采用大步长加快探索，非自由区域及路径收敛阶段采用小步长提升算法搜索精度，加快两棵树连接；最后，采用三次B样条曲线平滑规划路径，降低机械臂关节运动抖动，减少磨损，实现能耗优化。在三维环境中进行对比实验，以验证算法的优良性能。最终将融入机械臂避障功能后的算法部署到PUMA 560型机械臂，并在MATLAB R2022a仿真平台实现机械臂避障路径精准规划。更多还原

【Abstract】 It addresses the challenges of strong search randomness, slow convergence, and high computational complexity in robotic arm 3D path planning by proposing the Spatial Cell-Dynamic Target Guided Bounding box-Adaptive Step-BI-RRT~*,(SC-DTGB-AS-BI-RRT~*) algorithm. It combines adaptive step sizes with a Bidirectional Rapidly-exploring Random Tree(BI-RRT~*) approach, enhanced by space voxelization and dynamic target-guided bounding boxes.The algorithm′s multi-level optimization strategy first uses space voxelization to discretize 3D space, reducing computational load and boosting environmental perception accuracy. A dynamic target-guiding and bounding-box strategy adjusts the search focus across iterations: global exploration initially, a balance of exploration and convergence mid-way, and rapid path convergence later.The dynamic bounding box adapts the sampling range based on obstacle density, shrinking in cluttered areas and expanding in open spaces to enhance sampling efficiency. An adaptive step-size mechanism allows for large steps in free spaces and small steps near obstacles or during path convergence.Finally, cubic B-spline curves smooth the planned paths, reducing robotic arm joint jitter and improving energy efficiency. The algorithm′s effectiveness is validated through 3D comparative experiments. It is then integrated with a robotic arm′s obstacle-avoidance function and deployed on the PUMA 560 robotic arm. Accurate obstacle-avoidance trajectory planning is achieved using the MATLAB R2022a simulation platform.更多还原