摘要:
为应对复杂多变的环境，本文采用了体感技术和机器人控制技术与Kinect设备相结合的方法，设计了一个基于体感的机械手臂控制系统。为了更接近于人体手臂的结果，采用了六自由度的串联型机械手臂，使用51单片机作为主控板，Kinect Xbox One作为传感器，利用Kinect传感器无接触式采集人体动作信息并转化为控制指令，让机械手臂模仿并跟随人体运动。骨骼数据采集模块采集骨骼关节点的坐标，计算出关节角度并存入Mysql数据库中，Mysql数据库作为主要的通信桥梁，通过相似性算法将骨骼数据采集模块和机械臂控制模块结合起来，采集平台向数据库存入关节角度数据，控制平台从数据库中取出关节角度数据。控制模块计算出当前时刻与前一时刻关节角度的差值，然后与数据库中的数据匹配，从而识别运动轨迹，自动完成剩下的动作，也可以理解为机械手臂能做出“预判动作”。DTW算法的优势在于减小延时，使得机械手臂更加迅速地跟随人体手臂运动。实验结果证明该方法能有效地降低时滞性、平滑原始数据以及自动滤除异常数据。
Abstract:
In order to cope with the complex environment, this paper uses a combination of somatosensory technology and robot control technology and Kinect equipment to design a robotic control system based on somatosensory. In order to get closer to the results of the human arm, we used a six-degree-of-freedom series robot and used a 51 MCU as the main control board, and the Kinect Xbox One as a sensor. In order to make the robot arm imitate and follow the human movement, we used the Kinect sensor to collect human motion information without contact and convert it into control commands. The bone data acquisition module collects the coordinates of the bone joint points, calculates the joint angle and stores it in the Mysql database. As the main communication bridge, Mysql database combines the bone data acquisition module and the robot arm control module through the similarity algorithm. The acquisition platform enters joint angle data into the data inventory, and the control platform takes the joint angle data from the database. The control module calculates the difference between the joint angle of the current moment and the joint angle of the previous moment, and then matches it with the data in the database to identify the motion trajectory and automatically complete the remaining motions. It can also be understood that the robot arm can make a “pre-judgment action”. The advantage of the DTW algorithm is that it can reduce the delay so that the robot arm follows the movement of the human arm more quickly. Experimental results show that the method can effectively reduce the time lag, smooth the original data and automatically filter out the abnormal data.