An Energy-Efficient and Runtime-Reconfigurable FPGA-Based Accelerator for Robotic Localization Systems

A robot usually localizes itself in an environment by estimating the collection of its position and rotation states, while constructing a map of unknown surroundings, giving rise to the notion of Simultaneous Localization and Mapping (SLAM). SLAM is a fundamental kernel in autonomous machines at all computing scales, from drones, AR, VR to self-driving cars. Principled mathematical solutions for SLAM involve filtering-based or non-linear optimization-based (Fig. 1a), where the latter recently shows higher robustness but with intensive computation. Prior ASICs [1], [2] and FPGAs [3], [4], [5] have accelerated SLAM on hardware, but they usually target one specific design. In this work, we present a runtime-reconfigurable FPGA accelerator for robotic localization tasks. We exploit SLAM-specific data locality, sparsity, reuse, and parallelism, and achieve >5x performance improvement over the state-of-the-art. Especially, our design is reconfigurable at runtime according to the environment and platform to save power while sustaining accuracy and performance.