Study on odometry sensor alternative using 3D LiDAR for urban area application

Abstract

This paper presents experimental study on exploring Laser Odometry based LOAM for urban area applications. Most Odometry techniques to determine trajectories of vehicle in urban area use GPS, IMU, or Camera as sensing element. These sensors have their own weakness such as GPS prone to signal lost; IMU suffers from high drift error; and camera dependency to lighting conditions. Meanwhile, LOAM uses 3D LiDAR sensor and possess several advantages such as robust Odometry calculations and more resilient to lighting condition. LOAM was formulated using feature based detection to recognize feature point from edge line and planar surface inside the environment. These features are then used as reference points. The vehicle's position was estimated using dead reckoning method by comparing surrounding reference points with previously determined position. Performance evaluations are performed by comparing the recorded vehicle trajectory data against Google Cartographer which have centimetres accuracy. The result show that the approached strategy has comparable performance with Cartographer in urban area environment. It can achieve drift error between 0.6 and 10.0 metres during short distance travel (< 1.5 Km).