This research aims to improve autonomous navigation of coal mine rescue and detection robot, eliminate the danger for rescuers, and enhance the security of rescue work. The concept of model predictive control is introduced into path planning of rescue and detection robot in this paper. Sampling-Based Model Predictive Control (SBMPC) algorithm is proposed basing on the construction of cost function and predictive kinematics model. Firstly, input sampling is conducted in control variable space of robot motion in order to generate candidate path planning solutions. Then, robot attitude and position in future time, which are regarded as output variables of robot motion, can be calculated through predictive kinematics model and input sampling data. The optimum solution of path planning is obtained from candidate solutions through continuous moving optimization of the defined cost function. The effects of the three sampling methods (viz., uniform sampling, Halton’s sampling, and CVT sampling) on path planning performance are compared in simulations. Statistical analysis demonstrates that CVT sampling has the most uniform coverage in two-dimensional plane when sample amount is the same for three methods. Simulation results show that SBMPC algorithm is effective and feasible to plan a secure route for rescue and detection robot under complex environment.

中文翻译：

---

救援机器人路径规划的SBMPC算法研究

这项研究旨在改善煤矿救援和探测机器人的自主导航，消除对救援人员的危险，并提高救援工作的安全性。本文将模型预测控制的概念引入救援机器人的路径规划中。基于成本函数和预测运动学模型，提出了基于样本的模型预测控制算法。首先，在机器人运动的控制变量空间中进行输入采样，以生成候选路径规划解决方案。然后，可以通过预测运动学模型和输入采样数据来计算将来机器人的姿态和位置，这些姿态和位置被视为机器人运动的输出变量。路径规划的最佳解决方案是通过对定义的成本函数进行连续移动优化而从候选解决方案中获得的。在仿真中比较了三种采样方法（即均匀采样，霍尔顿采样和CVT采样）对路径规划性能的影响。统计分析表明，当三种方法的样本量相同时，CVT采样在二维平面上的覆盖范围最均匀。仿真结果表明，SBMPC算法在复杂环境下为救援机器人的安全路径规划是有效可行的。统计分析表明，当三种方法的样本量相同时，CVT采样在二维平面上的覆盖范围最均匀。仿真结果表明，SBMPC算法在复杂环境下为抢险侦察机器人规划安全路径是有效可行的。统计分析表明，当三种方法的样本量相同时，CVT采样在二维平面上的覆盖范围最均匀。仿真结果表明，SBMPC算法在复杂环境下为救援机器人的安全路径规划是有效可行的。