A novel autonomous lane navigation method for mobile robots

Wajahat Afsar, Purdue University

Abstract

Autonomous Guided Vehicles (AGVs) are a key area of research in mobile robotics. Autonomous lane detection and navigation is one of the main components of AGVs such as driverless cars. This research proposes an autonomous lane detection and navigation method for mobile robots based on detection of individual lane markers. First, an onboard camera is modeled, calibrated and distortions from camera are removed. A top-down view is generated. Depth sensor information is then used to extract the ground plane. Canny edge detection is used with contour detection to find lane markers, followed by estimation of lane midpoints. Random Sample Consensus is used to fit a lane model on the estimated midpoints. The lane model is refined using Catmull-Rom spline, a route is generated and the robot is navigated through the lane center. Experimental results on an indoor mobile robot demonstrate the robustness of the proposed method in navigating different lane conditions including various curved lanes as well as lanes with faded or lane markers.

Degree

M.S.E.

Advisors

Houshangi, Purdue University.

Subject Area

Engineering|Robotics|Computer science

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS