FPGA-Based Candidate Scoring Acceleration towards LiDAR Mapping

Cartographer is a system that can provide realtime simultaneous localization and mapping (SLAM) in both 2D and 3D across multiple platforms and sensors. Cartographer has been used in commercial applications such as Google Street View and in robotics applications such as interior mapping based on robot operating system (ROS) bag file recordings. However, cartographer is a complex system and incurs heavy computation demand. This paper aims at alleviating this demand by utilizing an FPGA-based approach to accelerate key components of Cartographer. Based on our analysis, we found that candidate scoring is the most time-consuming component of Cartographer. By implementing the candidate scoring function on an FPGA, the candidates can be scored at a much faster rate, which in turn allows Cartographer to generate a map much more quickly. In addition, our FPGA-based approach can lead to significant reduction in energy consumption when compared with traditional CPU-based approach, which makes it more suitable to be deployed in mobile and embedded computing scenarios.