Efficient Global 6D Localization in 3D TSDF Maps Using Point-wise and Scan-wise Reduction Methods on Embedded GPUs

Monte Carlo Localization is a widely used approach in the field of mobile robotics. While this problem has been well studied in the 2D case, global localization in 3D maps with six degrees of freedom has so far been too computationally demanding. Hence, no mobile robot system has yet been presented in the literature that can solve it in real-time. The computationally most intensive step is the evaluation of the sensor model, but it also offers high parallelization potential. In this paper, we investigate the massive parallelization of the evaluation of particles in truncated signed distance fields for three-dimensional laser scanners on embedded GPUs. In addition to providing a reference implementation, we compare two different parallelization schemes and compare the results to a CPU baseline. The results show that the GPU is up to 30 times faster and more than 50 times as energy efficient as the CPU implementation.