A new method for mobile robot arm blind grasping using ultrasonic sensors and Artificial Neural Networks

Abstract

The paper presents a new method to realize mobile robot arm grasping in indoor laboratory environments. This method adopts a blind strategy, which does not need the robot arms be mounted any kind sensors and avoid calculating the complex kinematic equations of the arms. The method includes: (a) two robot on-board ultrasonic sensors in base are utilized to measure the distances between the robot base and the front arm grasping tables; (b) an Artificial Neural Networks (ANN) is proposed to learn/establish the nonlinear relationship between the ultrasonic distances and the joint controlling values. After executing the training step using sampling data, the ANN can forecast/generate the next-step joint controlling values fast and accurately by inputting a new pair of real-time ultrasonic measured distances; (c) to let the blind strategy matching with the transportation process, an arm controlling component with user interfaces is developed; and (d) a method named training arm is adopted to prepare the training data for the training procedure of the ANN model. Finally, an experiment proves that the proposed strategy has good performance in both of the accuracy and the real-time computation, which can be applied to the real-time arm operations for the mobile robot transportation in laboratory automation.