Yiming GAN

Publications

^† indicates corresponding author; ^* indicates equal contribution.

Dedicated Accelerators for Robotic Applications

• Zheng Liu, Zeyu Guo, Zihan Liu, Anbang Wu, Han Zhao, Fangxin Liu, Zhezhi He, Yinhe Han, Jingwen Leng, Minyi Guo, Yiming Gan^†, Yu Feng, “Deltoris: Enabling Real-time VLA Inference in Embodied AI via Bit-level Sparsity and Speculative Inference,” 59th IEEE/ACM International Symposium on Microarchitecture (MICRO 2026).

• Yuhui Hao, Bo Yu, Shaoshan Liu, Yinhe Han, Yu Feng, Yiming Gan^†, “UDyr: An Adaptive Accelerator for High-Performance Humanoid Robot Control,” 59th IEEE/ACM International Symposium on Microarchitecture (MICRO 2026).

• Guoshuai Geng^*, Yuhui Hao^*, Yinhe Han, Yu Feng, Yiming Gan^†, “Velkoz: Generating Accelerators for Rigid-Flexible Robots through Domain Specific High-level Synthesis,” ACM Transactions on Design Automation of Electronic Systems (TODAES).

• Kaiyan Chang, Yiming Gan^†, Wenlong Zhu, Kun Wang, Zhirong Chen, Yuan Cheng, Yinhe Han, Huawei Li, Ying Wang, “ScaleGS: Closing the Gap between Real-time 3D Gaussian Splatting and Real-time XR Rendering,” ACM Transactions on Architecture and Code Optimization (TACO).

• Qiang Liu, Yihao Hua, Yuhui Hao, Bo Yu, Shaoshan Liu, Yiming Gan^†, “Unified and Efficient Factor Graph Accelerator Design for Robotic Optimization,” ACM Transactions on Architecture and Code Optimization (TACO).

• Yiyang Huang^*, Yuhui Hao^*, Bo Yu, Feng Yan, Yuxin Yang, Feng Min, Yinhe Han, Lin Ma, Shaoshan Liu, Qiang Liu, Yiming Gan^†, “Corki: Algorithm-Architecture Co-Design for Embodied AI-powered Robotic Manipulation,” 52nd International Symposium on Computer Architecture (ISCA 2025).

• Xinquan Lin, Haobo Xu, Yinhe Han, Yiming Gan^†, “HEX-SIM: Evaluating Multi-modal Large Language Models on Multi-chiplet NPUs,” IEEE International Symposium on Workload Characterization (IISWC 2024)  ·  Code.

• Yuhui Hao^*, Yiming Gan^*, Bo Yu, Qiang Liu, Yinhe Han, Zishen Wan, Shaoshan Liu, “ORIANNA: An Accelerator Generation Framework for Optimization-based Robotic Applications,” 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS 2024).

• Yuhui Hao^*, Yiming Gan^*, Bo Yu, Qiang Liu, Shaoshan Liu, Yuhao Zhu, “BLITZCRANK: Factor Graph Accelerator for Motion Planning,” 60th Design Automation Conference (DAC 2023).

• Qiang Liu, Yuhui Hao, Weizhuang Liu, Bo Yu, Yiming Gan, Jie Tang, Shaoshan Liu, Yuhao Zhu, “An Energy Efficient and Runtime Reconfigurable Accelerator for Robotic Localization,” IEEE Transactions on Computers (IEEE TC).

• Yu Feng, Gunnar Hammonds, Yiming Gan, Yuhao Zhu, “Crescent: Taming Memory Irregularities for Accelerating Deep Point Cloud Analytics,” 49th International Symposium on Computer Architecture (ISCA 2022).

• Weizhuang Liu^*, Bo Yu^*, Yiming Gan, Qiang Liu, Jie Tang, Shaoshan Liu, Yuhao Zhu, “Archytas: A Framework for Synthesizing and Dynamically Optimizing Accelerators for Robotic Navigation,” 54th IEEE/ACM International Symposium on Microarchitecture (MICRO 2021).

• Yiming Gan, Bo Yu, Boyuan Tian, Leimeng Xu, Wei Hu, Shaoshan Liu, Qiang Liu, Yanjun Zhang, Jie Tang, Yuhao Zhu, “Eudoxus: Characterizing and Accelerating Localization in Autonomous Machines,” 27th IEEE International Symposium on High-Performance Computer Architecture (HPCA 2021).

Robotic Applications and Algorithms

• Zixuan Wang, Huang Fang, Shaoan Wang, Yuanfei Luo, Heng Dong, Wei Li, Yiming Gan^†, “Hydra-Nav: Object Navigation via Adaptive Dual-Process Reasoning,” 43rd International Conference on Machine Learning (ICML 2026).

• Zixuan Wang, Bo Yu, Junzhe Zhao, Wenhao Sun, Sai Hou, Shuai Liang, Xing Hu, Yinhe Han, Yiming Gan^†, “KARMA: Augmenting Embodied AI Agents with Long-and-short Term Memory Systems,” IEEE International Conference on Robotics and Automation (ICRA 2025).

• Kai Pan, Yapeng Tian, Yinhe Han, Yiming Gan^†, “Benchmarking and Optimizing Federated Learning with Hardware-related Metrics,” 35th British Machine Vision Conference (BMVC 2024).

• Yiming Gan, Yuxian Qiu, Lele Chen, Jingwen Leng, Yuhao Zhu, “Low-Latency Proactive Continuous Vision,” 29th International Conference on Parallel Architectures and Compilation Techniques (PACT 2020) Best Paper Nomination.

• Igor Fedorov, Marko Stamenovic, Carl Jensen, Li-Chia Yang, Ari Mandell, Yiming Gan, Matthew Mattina, Paul N. Whatmough, “TinyLSTMs: Efficient Neural Speech Enhancement for Hearing Aids,” INTERSPEECH 2020.

• Yiming Gan, Yuxian Qiu, Jingwen Leng, Yuhao Zhu, “SVSoC: Speculative Vision Systems-on-a-Chip,” IEEE Computer Architecture Letters (CAL 2019).

Reliability and Fault Tolerance

• Wenhao Sun, Yuhui Hao, Yinhe Han, Yiming Gan^†, “Aphelios: A Selective Lock-step Neural Processing Unit Design,” 35th Great Lakes Symposium on VLSI (GLSVLSI 2025) Best Paper Nomination.

• Yiming Gan, Jingwen Leng, Bo Yu, Yuhao Zhu, “KINDRED: Heterogeneous Split-Lock Architecture for Safe Autonomous Machines,” ACM Transactions on Architecture and Code Optimization (TACO).

• Zishen Wan^*, Yiming Gan^*, Bo Yu, Shaoshan Liu, Arijit Raychowdhury, Yuhao Zhu, “The Vulnerability-Adaptive Protection Paradigm Toward Reliable Autonomous Machines,” Communications of the ACM (CACM 2024).

• Zishen Wan^*, Yiming Gan^*, Bo Yu, Arijit Raychowdhury, Yuhao Zhu, “VPP: The Vulnerability-Proportional Protection Paradigm Towards Reliable Autonomous Machines,” ISCA 2023 Workshop, 5th Domain Specific System Architecture (DOSSA-5).

• Abhishek Tyagi, Yiming Gan, Shaoshan Liu, Bo Yu, Paul Whatmough, Yuhao Zhu, “Thales: Formulating and Estimating Architectural Vulnerability Factor for DNN Accelerators,” 29th IEEE International Symposium on High-Performance Computer Architecture (HPCA 2023).

• Yiming Gan, Paul Whatmough, Jingwen Leng, Bo Yu, Shaoshan Liu, Yuhao Zhu, “BRAUM: Analyzing and Protecting Autonomous Machine Software Stack,” 33rd IEEE International Symposium on Software Reliability Engineering (ISSRE 2022).

• Yiming Gan^*, Yuxian Qiu^*, Jingwen Leng, Minyi Guo, Yuhao Zhu, “Ptolemy: Architecture Support for Robust Deep Learning,” 53rd IEEE/ACM International Symposium on Microarchitecture (MICRO 2020).

Thesis

• Yiming Gan, “Improving Fault Tolerance of Computing Systems for Autonomous Machines,” Ph.D. Dissertation, University of Rochester, 2023.

Research

I lead a research group at the Institute of Computing Technology, Chinese Academy of Sciences. We work on autonomous machines — an exciting area that may reshape the way we live. We hope to build autonomous machines that are more intelligent, more efficient, and more reliable. Our work spans new algorithms, new computer systems, and, hopefully, real robots.

AI for Chip Design

I believe the traditional chip design process is under a radical shift with the use of AI. With AI, we can move faster, while also needing to care more about the resilience of the design. Finding a way to efficiently use AI in chip design, quantify its ability, and improve that ability is extremely interesting to me.

• AI for Chip Design Arena (hiring)

• Resilient AI for Chip Design

Efficient Embodied AI Systems

I think robots will have the ability to think. The goal of this direction is to enable robots to accomplish long and complex tasks in real time. There are two fundamental challenges. First, although large language models have shown the ability to accomplish long-horizon tasks, the gap between the output of language models and the real physical world is huge. Second, the scale of large models makes it impossible for robots to finish their decision chain in real time. We currently have several related projects.

• Accelerator design for vision-language-action models

• Humanoid robot control (ongoing)

• Robotic GPU (ongoing)

• Hardwired action expert accelerator (ongoing)

Reliability

My research focuses on reliability, which I believe is particularly crucial in a world increasingly populated by robots. Ensuring reliability involves making software robust against various attacks and hardware resilient to different hazards. We are currently working on two related projects.

• Reliability for embodied AI systems (ongoing)

• Software–hardware co-design for fault tolerance in wafer-scale chips (hiring; prerequisite: basic computer architecture and compiler courses)

Team

We are a group of students and researchers building efficient and reliable computing systems for autonomous machines.

Ph.D. Students

• Guoshuai Geng — HLS for robotic applications (TODAES).

• Yuhui Hao — Accelerator design for humanoid robot control (TC, DAC'23, ASPLOS'24, ISCA'25, GLSVLSI'25, TACO'25, TODAES, MICRO'26).

• Mingyu Shi — Model-as-a-chip.

• Xuhui Qiu — RISC-V GPGPU.

Master's Students

• Yiyang Huang — Reliability and safety in embodied AI (ISCA'25).

• Zixuan Wang — Humanoid robot control (ICRA'25, ICML'26).

• Gaoyuan Liu — NPU virtualization.

• Jiale Wei — ROS 2 on RISC-V.

• Zhuomin Liu — RVV on RISC-V.

Alumni

• Wenhao Sun — Benchmarking real-time workloads (ICRA'25, GLSVLSI'25, JFR). Now Ph.D. student at HKUST.

• Sai Hou — Embodied AI robot system design (ICRA'25). Now Ph.D. student at Beijing Institute of Technology.

• Kai Pan — Benchmarking robotic applications (BMVC'24).

• Xinquan Lin — Multi-chiplet simulator design (IISWC'24). Now Ph.D. student at ICT, CAS.

Prospective Students

Much of my mentoring philosophy comes from my advisor. It has worked well between us, and I hope it will work well between us too. Generally, I am looking for students who are self-motivated, love to explore new things, and love to turn their findings into real systems. The ideal mentoring relationship should be interactive: I can help you with what I know, and you can expand my horizons with what you know.

What we are looking for

We welcome students from diverse backgrounds, even if your formal education did not involve extensive programming. It would be helpful if you are experienced in one or more of the following:

• Solid programming skills (Python, C++)

• Computer systems (computer architecture, FPGA, compilers, programming languages)

• Robotic applications (localization, planning, and control)

• Computer vision (deep learning, reinforcement learning, foundation models)

• Engineering mathematics (signal processing, optimization, matrix methods)

How to get in touch

I encourage you to read the research directions first, then send me an email. Please explain why your interests align with ours and how you can contribute. Attaching a CV, transcripts, and any code or papers you are proud of is always helpful.

• Email: ganyiming [at] ict.ac.cn