基于语义地图与稳定性评估的六足仿生机器人步态规划与协调控制

People and animals can go almost anywhere on Earth, which motivates the development of robot vehicles that use legs for their locomotion. The goal is to achieve animal-like mobility on rough and rugged terrain. A lot of engineers and scientists have embraced the opportunity of legged locomotion, building a diverse set of ingenious and inspiring legged robots, such as bipedal one, quadruped one, hexapod one, eight-legged one. The hexapod robot has higher stability than the quadruped one, and it is easier to coordinated control than the eight-legged one, and the motion is flexible and the reliability is high. So the hexapod robot occupies an irreplaceable position, especially on rough and rugged terrain. But, Six legged robot walking is still too mechanical to move very flexibly, and terrain adaptability is poor in the unstructured environment, which shows that it is not a good solution to discuss gait planning, trajectory planning or stability independently in order to get the desired result. Semantic map and stability evaluation-based gait planning and coordinated control are studied for a hexapod bionic robot whose single foot is 3-DOF and whose six legs is elliptical in distribution along the boy. Multi-posture dynamic stability analysis method, adaptive gait strategy, efficient multiped coordinated control method, route availability evaluation method is discussed one by one in the specific terrain. The environment information and the working posture of the robot are combined with dynamic stability, and the dynamic stability control algorithm is proposed. Visual perception and body stability are integrated into motion control. The deep semantic map and virtual scene model are used to calculate path passing probability with traversability assessment convolutional neural network, present a adaptive gait planning method with perception and anti-disturbance, and improve body motion planning and multiped coordination, which respond to better foot-end trajectory and body motion path. The project involves the terrain/pose dynamics model, the dynamic stability control, adaptive gait, multiped coordinated control, traversability assessment, and it is the goal that semantic map and stability evaluation-based gait planning and coordinated control theory and method is established for the mentioned hexapod robot. The results are expected to further promote the practical process of the multiped robot.

六足机器人仍存在行走过于机械、不能灵活运动、地形适应能力欠佳，而且非结构化环境决定了单一讨论机器人步态、轨迹或稳定性都难以获得理想的结果，本项目基于语义地图与稳定性评估研究三自由度单足机体椭圆布置的六足仿生机器人步态规划与协调控制，探索其特定地形下多位姿动稳定性分析方法、步态自适应策略、高效多足协调控制方法以及路线可通过性评估方法。将环境信息与机器人工况综合考虑进动稳定性判据，设计更适合六足机器人的动稳定性控制算法。将视觉感知和机身稳定性控制融入运动控制过程中，引入深层语义地图和虚拟模型，基于当前场景建立可通行性评估网络，实现有感知、抗扰动的自适应步态，改进机身运动规划和多足协调性，得到更优的足端运动轨迹和机体运动路径。本项目旨在建立结合语义地图与稳定性评估的六足机器人自适应步态规划与多足协调控制的理论与方法，进一步推进足类仿生机器人实用化进程。

六足机器人相较于其他轮式移动机器人，在非结构环境下具有较强的地形适应能力和良好的运动稳定性。本项目基于语义地图与稳定性评估研究三自由度单足机体椭圆布置的仿生六足机器人步态规划与协调控制，探索其特定地形下多位姿动稳定性分析方法、步态自适应策略、高效多足协调控制方法。. 在运动控制方面，为研究机器人机身与单腿之间的作用原理，设计了“3+3”构型测试台架，并提出了一种导纳控制策略，实现测试平台对单腿的力载荷功能，模拟出真实机身对单腿的力作用；为研究驱动系统与机器人机身之间的作用机制，基于虚拟模型控制在六足机器人上设置虚拟机构，通过其产生的虚拟力驱动机器人按照期望的轨迹运动，对机器人进行运动控制。为得到机器人的最优步态规划，通过对自然界足腿生物行走方式的研究，分析其在不同地形中的步态模式，规划六足机器人面对不同地形的运动步态，并进行稳定性评估，同时根据环境所获得的地形语义信息，使机器人在不同地形时切换至最优步态参数进行运动。. 在自主运动规划方面，设计了一种将随机抽样一致性光流跟踪法和改进的特征点匹配法结合的视觉里程计，解决了光流跟踪法定位精度不足、误差累积和特征点匹配法耗时久的问题；提出一种基于非线性优化的改进双目VI-SLAM算法，解决原有算法初始化阶段惯性参数估计不准确的问题。为提高机器人位姿估计的准确性，在前期激光点云实时语义分割方法基础上融合现有激光SLAM系统，构建语义辅助的激光SLAM系统；提出了改进的路径优化算法，进一步优化了机器人的运动路线。. 在复杂地形环境识别方面，为得到实时准确的语义地图，基于点云数据提出一种基于点和锥形栅格交互的激光点云实时语义分割方法，解决激光点云的稀疏性和密度不一致性所带的分割困难问题。. 上述研究成果在某种程度上有效促进了足式机器人相关领域的发展，进一步推动了足类机器人的实用化进程。