Signal Processing for Sensor Based Navigation of Mobile Robot

Sammanfattning: A self-navigating, path following and obstacle avoiding mobile robot is difficult to realize especially when its sensors are strongly effected by noise. This MSc thesis is aimed at investigating a realistic scenario of an autonomous mobile robot simulated in the MatLab environment. The robot system is able to follow a given reference path by utilizing its onboard sensors and decision making capabilities to avoid collisions with arbitrarily placed obstacles along its path. A novel navigational algorithm based on modifying the robot’s way-points using the run-time sensory data is developed and used to go around obstacles and then rejoin the original travel path as needed. The thesis work explores the impact of varying noise in the sensory data and ways of improving the navigational accuracy via signal processing. The study is done in two major sections, the first focusing on the navigational aspects of the mobile robot and the second exploring the sensory data analyses issues. The robot considered has a triangular shape with two differentially driven wheels at the rear left and rear right corners for skid steering control and one castor wheel in the front corner for balance purposes. The sensing system of the mobile robot includes infrared range finders with viewing angles of 180 degrees placed on the corners of the robot, which are able to detect obstacles all around the robot allowing effective path planning to be carried out via the special-purpose developed navigational algorithms. A reference path in an obstacle cluttered environment is assumed to be available for the robot to follow while avoiding randomly placed obstacles as the two wheels are driven to navigate the robot along the path using the robot kinematics. For making the navigation mobility of the robot as realistic as possible, practical infrared sensors have been studied experimentally to determine their noise characteristics to include in the simulation studies and the noise levels easily varied to simulate low and high noise levels and assess their effects on the overall navigational precision. Signal processing methods are used to show that improvements in the navigational performance can be achieved when the noise levels are high.