Design And Development Of Flexible Rapid Feed System For Robot

Nowadays,with the rapid development of the industrial manufacturing industry,assembly cost has become one of the main production costs for most manufacturing companies.How to save labor costs and improve production efficiency in the assembly process is a focused issue for manufacturing companies.Feeding is an important part of assembly,and traditional feeding is generally carried out manually.With the development of technology,manual feeding has been replaced by automatic feeding.The robot flexible rapid feed system is an automatic feeding device that changes the posture of the material through vibration and other methods and uses a robot to grasp the material to a set position.It is generally composed of a vibration feeding system,a visual recognition system,a control system,and a robot grasping system,The appropriateness of the vibration parameters in the vibration separation system directly affects the feeding speed of the feeding system.Compared to manual feeding,the robot flexible and fast feeding system has a faster feeding speed.However,the existing robot flexible rapid feed system uses human resources to conduct vibration parameter debugging one by one,resulting in problems such as labor consumption,low efficiency,and low flexibility,and it is unable to meet the factory requirements of short-term large batches of robot flexible feed systems with different specifications.Therefore,it is necessary to study how to improve the speed and flexibility of vibration parameter optimization in a robot flexible rapid feed system.Firstly,the hardware of the robot flexible rapid feed system.is constructed in this thesis.The voice coil motor is used as the vibration source of the feeding system,suitable motor drivers are selected to regulate the voice coil motor,and PLC is used to control the vibration parameters of all voice coil motors.In order to better identify the posture of the material,the light source,industrial camera,and lens are selected,and the overall route of the robot flexible and rapid feed system is connected.The software and hardware connectivity of the feeding system and communication between various software are completed,achieving the vibration function of the robot flexible rapid feed system.Secondly,this article incorporates reinforcement learning into the optimization of vibration parameters of the robot flexible fast feeding system.Based on the inherent characteristics of the robot flexible feeding system,an improvement is made to the SARSA algorithm in reinforcement learning,setting the expected total number of steps per scenario,and feeding back the number of materials that meet the grasping conditions obtained after each system vibration to the reward function of the algorithmε-Optimization of SARSA algorithm in greedy functions,Through experimental debugging,the flexible feeding system of the robot can obtain more satisfying conditions to grab materials after a certain number of vibrations,avoiding manual vibration parameter debugging one by one.This enables the flexible and fast feeding system of the robot to have selflearning ability in vibration parameter debugging,reducing manual participation in vibration parameter debugging of the feeding system.Compared to the traditional SARSA algorithm,the improved SARSA algorithm proposed in this thesis has a faster speed in optimizing the vibration parameters of the feeding system.In order to further improve the optimization efficiency and work efficiency of the vibration parameters of the robot flexible feeding system,this article combines the position information of the materials to improve the Q-learning algorithm in reinforcement learning.The QPD(Q-learning Position Distance)algorithm and QPN(Q-learning position number)algorithm are proposed,which add the position information of the materials after the system vibration to the reward function and optimize them respectively,Calculate the greedy value based on the number of items that meet the grabbing conditions in each scenario ε Make adjustments to enable the flexible and fast feeding system of the robot to obtain more materials that meet the grasping conditions with fewer vibrations,and use reinforcement learning algorithms to control the position of the materials.The vibration parameters optimized by the QPD algorithm and QPN algorithm are vibrated to obtain materials that meet the gripping conditions.The average distance between the materials and the starting point pixel of the robotic arm is less than the longest average pixel distance,accounting for 24.35% and 26.07% of the total vibration times in the experiment,which is more than 7.8% of the traditional Q-learning algorithm.This indicates that the probability of QPD algorithm and QPN algorithm obtaining materials that meet the gripping conditions being closer to the starting point of the robotic arm is higher,This reduces the movement time of the robotic arm and increases the efficiency of the system.Finally,this thesis summarizes the research done and prospects the future research direction of robot flexible rapid feed system.