基于连续隐半马尔可夫模型的复杂轨迹演示示教算法研究
Research on Teaching Algorithm of Complex Trajectory Demonstration Based on Continuous Hidden Semi-Markov Model

针对工业机器人传统示教方法示教效率低、操作难度大的问题，提出了一种演示示教算法，通过人对目标轨迹的多次演示生成形状复杂且位置精确的示教轨迹。将轨迹中运动状态发生较大变化的点定义为关键点，使用改进的多尺度曲率积算法和连续隐半马尔可夫模型获取不同示教轨迹特征一致的公共关键点。以各个公共关键点簇的中心点作为示教轨迹分割点，将形状复杂、不易拟合的轨迹分割成多条结构简单的子轨迹，利用最小二乘B样条分段拟合曲线以形成最终的示教轨迹。通过实验证明该算法具有良好的示教精度及易用性。
Aiming at the problems of low teaching efficiency and difficult operation in the traditional teaching method of industrial robots, a demonstration teaching algorithm is proposed, which generates a teaching trajectory with complex shape and accurate position through multiple demonstrations of the target trajectory by humans. The points in the trajectory with large changes in motion state are defined as key points, and the improved multi-scale curvature product algorithm and continuous hidden semi-Markov model are used to obtain common key points with consistent characteristics in different taught trajectories. Taking the center point of each common key point cluster as the teaching trajectory segmentation point, the complex-shaped and difficult-to-fit trajectory is divided into simple-shaped sub-trajectories, and the least squares B-spline is used to segmentally fit the curve to form the final teaching trajectory. Experiments show that the algorithm has good teaching accuracy and ease of use.