Laser Speckle Velocimeter for Mobile Industrial Robot Odometry

This paper reports the design and construction of a laser speckle velocimeter for application to mobile industrial robots. It is intended to supplement the typically used simultaneous localisation and mapping (SLAM) approach by providing enhanced odometry that unlike wheel encoders is unaffected by wheel slippage and will improve navigation in feature-sparse or varying environments. The sensor was installed on an industrial mobile robot (MIR200) and performance trials conducted in a manufacturing environment. We report a novel signal processing scheme that was developed for reliable operation across a range of surface features typically found in these environments, and present the results of odometry benchmarks conducted in a real-world environment. The odometry calculated using the laser speckle sensor was assessed in comparison to the MIR200's onboard SLAM implementation and to ground truth measurements of robot position using a laser tracker. The overall performance of odometry calculated using the speckle sensor and robot heading was comparable to that achieved by the robot's SLAM system, showing odometry errors of 0.07-0.46% over a 48 m path. Over the outward path (the first 24 m) the speckle-based odometry performed better than SLAM, with errors of 0.07-0.46% for the speckle and 0.16-1.22% for the robot's SLAM algorithm. It is therefore anticipated that further performance improvements could be made by fully integrating the sensor output and SLAM data.