Autonomous Obstacle Avoidance Robot Using Regression

EasyChair Preprint no. 3581

Abstract

Obstacle avoidance is considered as one of the main features of autonomous intelligent systems. There are various methods for obstacle avoidance. In this paper, obstacle avoidance is achieved by the difference between left wheel velocity and right wheel velocity of differential drive robot. The magnitude of difference between the wheel velocities is used to steer the robot in correct direction. Data is collected by driving the robot manually. Ultrasonic sensors are used for distance measurement and IR sensors are used to collect the data of wheel velocities. This data is used to build a linear machine learning model which uses sonar data as input features. The model is used to predict the wheel velocities of the differential drive robot. The model built is then programmed into Atmega328 microcontroller using Arduino IDE. This enables the mobile robot to steer itself to avoid the obstacles. Since all the components used for this robot are highly available and cost effective, the robot is economically affordable.

Keyphrases: Arduino, autonomous mobile robot, linear regression, NodeMCU, obstacle avoidance, pseudo-inverse, Raspberry Pi, stochastic gradient descent

@Booklet{EasyChair:3581,
  author = {Naveen Vakada and Aasish Chunduri and Kavya Manne and Vidhya Lakshmi Meda and Kl Sailaja},
  title = {Autonomous Obstacle Avoidance Robot Using Regression},
  howpublished = {EasyChair Preprint no. 3581},

  year = {EasyChair, 2020}}