An FPGA Acceleration and Optimization Techniques for 2D LiDAR SLAM Algorithm
An FPGA Acceleration and Optimization Techniques for 2D LiDAR SLAM Algorithm
复制标题
2D LiDAR SLAM 算法的 FPGA 加速和优化技术
DOI：
10.1587/transinf.2020edp7174
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Hiroki Matsutani
中科院分区：
文献类型：
--
作者：
K. Sugiura;Hiroki Matsutani
对于计算资源有限的自主移动机器人来说，高效的同时定位与映射（SLAM）方法硬件设计是必不可少的。在本文中，我们开发了一种资源高效的FPGA设计，用于加速扫描匹配过程，这通常是二维激光雷达SLAM方法的瓶颈。扫描匹配是通过将最新的激光雷达测量结果与占用网格地图（对周围环境信息进行编码）对齐来纠正机器人姿势的过程。该设计利用基于Rao-Blackwellized Particle Filter （RBPF）算法的固有并行性，对多个粒子并行执行扫描匹配计算。在设计中，采用映射压缩技术和查找表技术来降低资源的利用率，实现最大的吞吐量。使用基准数据集的仿真结果表明，在不严重降低最终输出质量的情况下，扫描匹配速度提高了23.3-51.1倍，总体吞吐量提高了1.97-3.16倍。此外，我们的实现只需要Xilinx ZCU104评估板中可用资源的37%，从而为实现SLAM在室内移动机器人上的应用提供了可行的解决方案。
An efficient hardware design of Simultaneous Localization and Mapping (SLAM) methods is of necessity for mobile autonomous robots with limited computational resources. In this paper, we develop a resource-efficient FPGA design for accelerating the scan matching process, which typically exhibits the bottleneck in 2D LiDAR SLAM methods. Scan matching is a process of correcting a robot pose by aligning the latest LiDAR measurements with an occupancy grid map, which encodes the information about the surrounding environment. The proposed design exploits an inherent parallelism in the Rao-Blackwellized Particle Filter (RBPF) based algorithms to perform scan matching computations for multiple particles in parallel. In the design, map compression technique and lookup-table are employed to reduce the resource utilization and achieve the maximum throughput. Simulation results using the benchmark datasets show that the scan matching is accelerated by 23.3-51.1x and the overall throughput is improved by 1.97-3.16x without seriously degrading the quality of the final outputs. Furthermore, our implementation requires only 37% of the total resources available in the Xilinx ZCU104 evaluation board, thus providing a feasible solution to realize SLAM applications on indoor mobile robots.