Abstract

Computer vision-based path planning can play a crucial role in numerous technologically driven smart applications. The proposed computer vision-neural network technique was fed by two photographs that were taken from two separate vantage points in order to get accurate spatial coordinates of objects in real time. These photographs were shot from various angles. After that, Q-learning was utilized to build a set of fundamental operations, such as up, down, left, right, backward, and forward. These basic operations include moving up, down, left, and right, as well as forward and backward. In the end, a neural network that had been trained was used to find a set of joint angles that related to the activities that were detected. These angles were found by correlating the activities that were recognized with the joint angles.