Simulation of Autonomous Multifunctional Mobile Robot Using Machine Vision

In the recent decade, vision-based robotic systems are employed for numerous domestic and industrial applications. Currently, a majority of mobile robots lack multiple capabilities and are confined to perform specific tasks due to the drawbacks of conventional distance sensors. This paper demonstrates the integration of road sign recognition, leader-follower, object tracking and self-driving functionalities into a single multifunctional robot using machine vision. The proposed object tracking algorithm uses dimensions of the bounding box enclosing the object and location of the object's centre in the recognition frame to generate the necessary motion commands for the robot. A 2-D convolutional neural network is developed to predict precise steering angles associated with the captured images for end-to-end steering control in the simulation environment. The simulation results obtained in Webots and Udacity-self driving car simulator platforms prove the robot's adaptability for multiple applications