Research On Mobile Robot Localization Based On Inertial Navigation And Visual Sensor Information Fusion

A reliable localization is the basic and the most important function for mobile robot. Andit becomes one of the most remarkable and the most challenge research projects recent years.This paper studied the problem of robot posture tracking technology in a known andstructured circumstance, and we applies a multi-sensor fusion algorithm to integratedpositioning system, which is based on inertial navigation and robotic vision. The maincontents include the following aspects:Firstly, we research the inertial navigation location. firstly we analyses the principle ofodometer and gyroscope, and then according to the motion feature of the mobile robot, webuild kinematics model of two-wheeled differential mobile robot. And using discrete kalmanfilter, we accomplishes the data fusion between odometer and gyroscope. Through thesimulation experiment, the paper contrasts the track deviation about the state of before andafter fusion.The second, we research the visual location. At first, the paper introduces the cameramodel with imaging principles including proposed the imaging model of camera and therelationship between camera imaging and four coordinate systems. Then, it calibrates thecamera characteristics, introduces the hardware and software structure of robot visual locationsystem, analyses the process of localization and studies the location technology of robot baseon POSIT algorithm.The third, the paper studies three kind of robot location algorithm base on multi-sensorinformation fusion in term of EKF algorithm, Hybrid Location Algorithm and Weighted SumAlgorithm. It realizes robot localization based on the data fusion between inertial navigationand visual sensor.Finally, using the MATLAB simulations, we compares the three kinds of locationconsequence in the previous chapter, it is proved that the EKF is the best fitting for robotlocalization by comparing the result with track plotting only with a odometer and calculatingthe position of localization with fusion the data between discrete kalman filter，odometer andgyroscope. Comparative analysis and the simulation results from the two aspects of theposition error and orientation error, The study found that the EKF position error is smaller, thebetter the positioning of the simulation approach goal.