DRACO: Co-design for DSP-Efficient Rigid Body Dynamics Accelerator

TL;DR

A precision-aware quantization framework that reduces DSP demand while preserving motion accuracy while preserving motion accuracy is proposed and an inter-module DSP reuse methodology to improve DSP utilization and save DSP usage is presented.

Abstract

We propose a hardware-efficient RBD accelerator based on FPGA, introducing three key innovations. First, we propose a precision-aware quantization framework that reduces DSP demand while preserving motion accuracy. This is also the first study to systematically evaluate quantization impact on robot control and motion for hardware acceleration. Second, we leverage a division deferring optimization in mass matrix inversion algorithm, which decouples reciprocal operations from the longest latency path to improve the performance. Finally, we present an inter-module DSP reuse methodology to improve DSP utilization and save DSP usage. Experiment results show that our work achieves up to 8x throughput improvement and 7.4x latency reduction over state-of-the-art RBD accelerators across various robot types, demonstrating its effectiveness and scalability for high-DOF robotic systems.

Figures (13)

• Figure 1: Robotics pipeline and motion planning & control framework.
• Figure 2: Overview of the motivation, challenges and contributions.
• Figure 3: RBD functions and current architecture. (a) RBD functions overview. (b) Round Trip Pipeline architecture in DaduRBD. (c) Architecture for accelerating RBD functions.
• Figure 4: Overview of the quantization framework.
• Figure 5: Error propagation and error compensation. (a) Forward pass of RNEA algorithm. (b) Quantization error of RNEA's fpass. (c) Velocity quantization error of joints of iiwa robot. (d) Average absolute error of quantized $M^{-1}$ matrix before and after compensation. (Small errors are not shown in this figure.)