3D SLAM Using Voxel Grid Filter on Wheeled Mobile Robot

This study introduces an innovative approach to localization and mapping field by integrating sensor fusion into a non-holonomic differential mobile robot. The suggested approach harmoniously combines real-time Simultaneous Localization and Mapping with the Voxel Grid Filter to downsample point cloud data, achieving an ideal balance between generating dimensionally accurate maps and adhering to the system's computational constraints. Instead of processing each individual data point, the system calculates a single representative point for each voxel by aggregating adjacent points into larger voxel units (volumetric pixels) using the centroid or average position of all points within that voxel. The system efficiently minimizes computational demands while maintaining a representative surrounding model through voxel-based points. Simulation results demonstrate the efficacy of this approach in tackling SLAM-related challenges, particularly in terms of map accuracy and computational efficiency. Utilizing a 0.01 voxel grid filter notably improves iteration efficiency, reducing the average computation time per algorithm execution to 0.160 seconds. In contrast, processing raw LiDAR data requires 0.247 seconds per iteration, resulting in a 35% reduction in computational time. This enhancement in efficiency is achieved without sacrificing the accuracy of the generated maps or the precision of robot localization, demonstrating the viability of this method for real-time SLAM applications.