烹饪机械手运动控制与机器学习防碰撞策略研究

Research on Motion Control of Cooking Manipulator and Machine Learning on Anti-Collision Strategy

为研究一种基于强化学习的协作机器人轨迹示教和烹饪学习方法,以轻量一体化双臂UR型机器人进行食品烹饪为实例,分析了机械臂的运动学正反解问题。首先,介绍了双臂型UR机器人厨房烹饪的控制方案,以及如何借助优化策略来提高主观食品质量评级;其次,研究了利用机器人运动学解决了烹饪机械手运动控制问题。再次,讨论了如何利用在线强化学习模块进行拖拽示教和碰撞保护学习,以及如何利用基于向量机的碰撞检测方法保证烹饪过程中机器人左右手执行烹饪的安全性。最后对食品烹饪机器人将进行了实验验证。结论表明机器人的运动学算法和调制烹饪流程具备良好的准确性和有效性,能在可靠地提高食品烹饪质量的同时实现批量生产。

Robotic chef cooking is an effective way to reduce the cooking burden of chefs by learning human skills.In order to study a method of trajectory instruction and cooking learning for cooperative robots based on reinforcement learning,the forward and backward kinematics of a lightweight integrated dual-arm UR robot for food cooking was analyzed.Firstly,the cooking control scheme of the double-arm UR robot kitchen,and the method of improving the subjective food quality rating by means of optimization strategies was introduced;secondly,the problem of cooking manipulator motion control was solved by using robotic kinematics.Thirdly,the online reinforcement learning module was used for drag teaching and collision protection learning,and the collision detection method based on vector machine was used to ensure the safety of the left and right hands of the robot to perform cooking during the cooking process.Finally,the food cooking robot was experimentally verified.The conclusion shows that the robot's kinematics algorithm and modulation cooking process have good accuracy and effectiveness,and can reliably improve the quality of food cooking while realizing mass production.