RoboPIM: A ReRAM-based Accelerator for LLM-based Robotics Applications via Dynamic Task Slicing
Xiao, Wenjing Wang, Jianyu Chen, Dan Li, Huize Guizani, Mohsen Chen, Min Wu, Thomas
Xiao, Wenjing
Wang, Jianyu
Chen, Dan
Li, Huize
Guizani, Mohsen
Chen, Min
Wu, Thomas
Supervisor
Department
Machine Learning
Embargo End Date
Type
Journal article
Date
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
1Large Language Models (LLMs) have demonstrated significant potential in improving robotics applications through natural language processing capabilities. Matrix multiplication dominates the computational workload in these LLM-based robotics systems, and Resistive Random Access Memory (ReRAM) provides notable performance benefits due to its inherent parallelism and energy efficiency. However, matrix multiplications in these applications pose dual heterogeneity challenges, encompassing both structural and scale heterogeneity. Structurally, these applications involve both standard dense matrix operations and block-structured sparse multiplications, where sparsity patterns vary significantly with robot topology. In terms of scale, matrix sizes range from just a few to thousands of dimensions, making large ReRAM crossbars designed for LLM inefficient for small-scale robotics computations. Existing acceleration approaches fail to address this dual heterogeneity, limiting their effectiveness for LLM-based robotics applications. In this work, we propose RoboPIM, the first low-energy ReRAM-based accelerator tailored for LLM-based robotics applications. RoboPIM adapts to the diverse matrix multiplication requirements across robot platforms and LLM configurations. Our design features a ReRAM architecture with various crossbar sizes and implements a dynamic task slicing scheme that allocates matrix slices to appropriately sized crossbars. To maximize ReRAM’s parallel processing capabilities, we develop a two-stage scheduling strategy that efficiently manages matrix multiplications while enhancing hardware utilization. Comprehensive evaluations demonstrate that RoboPIM achieves 2.85× and 6.85× speedup across diverse robotic platforms and LLMs respectively, outperforming state-of-the-art domain-specific solutions.
Citation
W. Xiao, J. Wang, D. Chen, H. Li, M. Guizani, M. Chen, T. Wu, "RoboPIM: A ReRAM-based Accelerator for LLM-based Robotics Applications via Dynamic Task Slicing," IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. PP, no. 99, pp. 1-1, 2026, https://doi.org/10.1109/tcad.2026.3660201.
Source
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Conference
Keywords
40 Engineering, 46 Information and Computing Sciences, 4608 Human-Centred Computing, 7 Affordable and Clean Energy
Subjects
Source
Publisher
IEEE