Robust and Reliable LiDAR Signal Filtering using Recursive Weighted Myriad-Based Filter
Khai-Hoong Khoo, Chuan-Hsian Pu, S. Anandan Shanmugam, Pei Cheng Ooi, Vimal Rau Aparow, Hermawan Nugroho · 2023
The echoes of the LiDAR signal could experience various sporadic reflections from the external clutters of the environment, this is particularly prominent in an indoor navigation environment. The sporadic echoes from various reflections with different reflection coefficients from different objects from adverse weather conditions, malfunctioning LiDAR or multiple echoes from different paths could cause unreliable estimation and incur excessive processing power using a microcontroller unit (MCU) for computation. In this work, we propose a robust recursive weighted myriad filtering (RWMyF) technique to remove the sporadic echoes from the LiDAR under two test cases, which are clean and contaminated signals. The contaminated LiDAR signals are generated by superimposing Gaussian noise on the originally obtained LiDAR scan data. Besides that, this work proposed a method to convert the LiDAR scanned data from its polar coordinates into 2D cartesian coordinates. From the results, the sporadic impulsive noise from the RPLiDAR detection could be effectively eliminated through the RWMyF. The filtered and reconstructed LiDAR graph, in both test cases of clean and contaminated signals in polar coordinates, could provide better and more reliable data to MCU to generate reliable distance estimates for collision avoidance. Hence, the removal of sporadic echoes is essential for obstacle avoidance in autonomous navigation applications.