Optimization of Work Place for Robotic Arms using Artificial Neural Network for Object Locator

Artificial Neural Network (ANN) is a powerful and one of the most effective learning techniques and is inspired by biological neural networks. This paper is focus on training a robotic arm to accurately locate the object within the arm's range. This will be accomplished by creating a system with the implementation of ANN the object's location will be fed as an input to the neural network. The neural network will process this data such that output will be a set of two angles at which each joint angle of the robotic arm should move to accurately locate the aforementioned object. Methodology focuses on train a robotic arm to accurately locate and pick an object within the arm's range and analysis of error within the workplace to optimize the results. This will be accomplished by creating a system with the implementation of ANN. The performance of the system is measured by calculating the error or the difference between the output angles of the neural network and the measured angles. Although there are no straight forward relationships between the input and outputs, the neural network was able to fit reasonable output angles with corresponding input coordinates to properly locate the object. The neural network nonlinearly predicted the appropriate outputs for a given set of inputs. The outputs of the neural network as it was fed with the input training data are compared against the output training data to measure its accuracy. The designed ANN was able to learn the relationship between the input coordinates and the output angles as shown by the results. However, there are some inputs to the neural network that it cannot accurately process to give a reasonable output. This can be improved by adding more training data to the system so that the calculated weights by the neural network can be corrected. These additional training data can help the neural network to decide on which weights to converge to will be correct. Also, adding more hidden layers and more neurons to each layer can help solve this. The normalization of Distance from robotic arm data for the trained neural network using the sensors input can be implemented for analysis of workplace to improve the performance of robotic arm. Keywords - Artificial Neural Network (ANN), Back Propagation Neural Network (BPNN), Robotic Arm, Learning Process, Object Locator.