pith. sign in

arxiv: 2605.24044 · v1 · pith:CMDVNP4Nnew · submitted 2026-05-21 · 💻 cs.RO · cs.SE· cs.SY· eess.SY

RED: Adaptive Real-Time DAG Scheduling for Robotic Inference under Environmental Dynamics

Zexin Li , Tao Ren , Johnathan Liu , Xiaoxi He , Cong Liu This is my paper

Pith reviewed 2026-06-30 15:58 UTC · model grok-4.3

classification 💻 cs.RO cs.SEcs.SYeess.SY
keywords real-time schedulingDAG schedulingrobotic inferenceenvironmental dynamicsmulti-task DNNMIMONetdeadline guaranteesresource-constrained platforms
0
0 comments X

The pith

A deadline-aware scheduler for robotic multi-task inference adapts to environmental changes while preserving end-to-end timing guarantees.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Robots in dynamic environments face changes that reshape their computation tasks at runtime, such as new tasks appearing or precedence relations shifting. This can degrade performance for multi-task deep neural network inference under tight real-time budgets. RED is a scheduling framework designed to adapt to these Robotic Environmental Dynamics on resource-constrained platforms. It does so by using a deadline-aware scheduler that assigns intermediate sub-deadlines to handle evolving graphs and asynchronous inference. The approach also refines MIMONet structures to align with schedulability needs, and experiments on hardware show gains in throughput and deadline satisfaction.

Core claim

RED is a real-time scheduling framework for multi-task deep neural network workloads on resource-constrained robotic platforms that adapts to Robotic Environmental Dynamics while preserving end-to-end timing guarantees under modeling assumptions. The core is a deadline-aware scheduler that assigns intermediate sub-deadlines to accommodate evolving computation graphs and asynchronous inference induced by unpredictable conditions. The framework supports flexible deployment of MIMONet through workload refinement and graph-reconstruction that aligns the structure with schedulability requirements.

What carries the argument

The deadline-aware scheduler assigning intermediate sub-deadlines to accommodate evolving computation graphs and asynchronous inference, combined with MIMONet workload refinement and graph-reconstruction procedure.

If this is right

  • Consistent gains in throughput over existing methods.
  • Improved deadline satisfaction rates.
  • Greater robustness to interference from environmental changes.
  • Better adaptability while maintaining real-time performance.
  • Lower runtime overhead on platforms like NVIDIA Jetson and Apple M-series.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Such adaptive scheduling could be applied to other embedded AI systems facing variable workloads, like autonomous vehicles or smart sensors.
  • Hardware designers might optimize for shared-parameter networks like MIMONet to leverage these refinements.
  • Further tests could explore scaling to multi-robot coordination scenarios with shared resources.

Load-bearing premise

The modeling assumptions hold such that the deadline-aware scheduler can preserve end-to-end timing guarantees even as computation graphs evolve and inference becomes asynchronous due to unpredictable conditions.

What would settle it

A test run on the evaluated platforms where, despite the modeling assumptions, the system misses end-to-end deadlines when the workload structure changes due to simulated environmental dynamics.

Figures

Figures reproduced from arXiv: 2605.24044 by Cong Liu, Johnathan Liu, Tao Ren, Xiaoxi He, Zexin Li.

Figure 1
Figure 1. Figure 1: Environmental changes reshape the workload DAG of a running robot. As dynamics [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: MIMONet versus a traditional single-input/single-output (SISO) DNN. The shared en [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Asynchrony amplifies timing risk. We compare response-time distributions and task out [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: High-level architecture of RED. The framework blends intermediate-deadline scheduling, [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Applying each RED component on a MIMONet-derived DAG. [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Multi-DAG optimizations and control flow in RED. [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Overall effectiveness of RED evaluated on four different resource-constrained intelligent [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: A practical case study on ROS2 in￾volving a camera input with multiple inference outputs on Xavier. frameworks such as S3DNN [114] and ApNet (RTSS’18) is not applicable: those systems target static, single-DNN inference pipelines and do not support MIMONet-style multi-input multi-output structures or graph mutation, so adapting them as baselines would require re-architecting their core dispatch logic and w… view at source ↗
read the original abstract

Robots deployed in dynamic environments must contend with environment-driven changes that reshape computation at runtime: new tasks may appear, precedence relations can shift, and overall workload structure evolves, all of which degrade performance, especially when multi-task inference is required under tight resource and real-time budgets. We present RED, a real-time scheduling framework for multi-task deep neural network workloads on resource-constrained robotic platforms that adapts to Robotic Environmental Dynamics (RED) while preserving end-to-end timing guarantees under modeling assumptions. The core of RED is a deadline-aware scheduler that assigns intermediate sub-deadlines, allowing it to accommodate evolving computation graphs and asynchronous inference induced by unpredictable conditions. The framework also supports flexible deployment of MIMONet (multi-input multi-output neural networks), commonly used in multi-tasking robots to alleviate memory pressure through weight sharing. RED explicitly leverages this shared-parameter property via a workload refinement and graph-reconstruction procedure that aligns MIMONet structure with schedulability requirements, improving compatibility and efficiency. We implement RED on NVIDIA Jetson family platforms and on an Apple M-series MacBook and evaluate it on navigation-oriented workloads representative of real robotic scenarios. Experiments show consistent gains over existing methods in throughput, deadline satisfaction, robustness to interference, adaptability, and runtime overhead.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The paper presents RED, a real-time scheduling framework for multi-task deep neural network workloads on resource-constrained robotic platforms. It adapts to Robotic Environmental Dynamics by accommodating evolving computation graphs and asynchronous inference via a deadline-aware scheduler that assigns intermediate sub-deadlines, while preserving end-to-end timing guarantees under modeling assumptions. The framework supports MIMONet deployment through workload refinement and graph reconstruction to align with schedulability needs, and is implemented on NVIDIA Jetson platforms and Apple M-series hardware with evaluations on navigation-oriented robotic workloads showing gains in throughput, deadline satisfaction, robustness to interference, adaptability, and runtime overhead.

Significance. If the modeling assumptions prove realistic for robotic scenarios and the experimental results are reproducible with appropriate statistical validation, the work could meaningfully advance adaptive real-time scheduling for dynamic multi-task inference in robotics, a practical area where existing methods often struggle with evolving DAGs and resource constraints.

minor comments (2)
  1. [Abstract] Abstract: The acronym RED is overloaded, referring both to the proposed framework and to 'Robotic Environmental Dynamics'; this risks confusion and should be disambiguated in the title, abstract, and introduction.
  2. [Abstract] Abstract: Claims of 'consistent gains' and 'improving compatibility and efficiency' are stated without quantitative values, baselines, or metrics; the full manuscript should supply these details for evaluation.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their review of our manuscript on RED. We appreciate the recognition of the framework's potential to advance adaptive real-time scheduling for dynamic multi-task inference in robotics. The recommendation is listed as uncertain, but no specific major comments are enumerated in the report. We therefore provide no point-by-point responses below and stand ready to address any additional points the referee may wish to raise.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The abstract and description introduce RED as a scheduling framework using a deadline-aware scheduler for evolving DAGs and MIMONet refinement, with timing guarantees scoped explicitly to modeling assumptions. No equations, parameter fits, self-citations, or derivations are presented that reduce any claimed prediction or result to its inputs by construction. The framework is described at a high level without self-definitional loops or renamed known results. This matches the reader's assessment of no visible circularity signals, making the derivation self-contained against the supplied text.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; the timing guarantees rest on unspecified modeling assumptions.

pith-pipeline@v0.9.1-grok · 5767 in / 1001 out tokens · 35194 ms · 2026-06-30T15:58:45.643360+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

116 extracted references · 20 canonical work pages · 5 internal anchors

  1. [1]

    Com- mongraph: Graph analytics on evolving data

    Mahbod Afarin, Chao Gao, Shafiur Rahman, Nael Abu-Ghazaleh, and Rajiv Gupta. Com- mongraph: Graph analytics on evolving data. InProceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Vol- ume 2, ASPLOS 2023, page 133–145, New York, NY, USA, 2023. Association for Computing Machinery

    2023

  2. [2]

    Com- mongraph: Graph analytics on evolving data (abstract)

    Mahbod Afarin, Chao Gao, Shafiur Rahman, Nael Abu-Ghazaleh, and Rajiv Gupta. Com- mongraph: Graph analytics on evolving data (abstract). InProceedings of the 2023 ACM Workshop on Highlights of Parallel Computing, HOPC ’23, page 1–2, New York, NY, USA,

    2023

  3. [3]

    Association for Computing Machinery

  4. [4]

    Exact response-time bounds of periodic dag tasks under server-based global scheduling

    Shareef Ahmed and James H Anderson. Exact response-time bounds of periodic dag tasks under server-based global scheduling. In2022 IEEE Real-Time Systems Symposium (RTSS), pages 447–459. IEEE, 2022

    2022

  5. [5]

    The federated scheduling of systems of conditional sporadic dag tasks

    Sanjoy Baruah. The federated scheduling of systems of conditional sporadic dag tasks. In 2015 International Conference on Embedded Software (EMSOFT), pages 1–10. IEEE, 2015

    2015

  6. [6]

    Predjoule: A timing-predictable energy optimization framework for deep neural networks

    Soroush Bateni, Husheng Zhou, Yuankun Zhu, and Cong Liu. Predjoule: A timing-predictable energy optimization framework for deep neural networks. In2018 IEEE Real-Time Systems Symposium (RTSS), pages 107–118. IEEE, 2018

    2018

  7. [7]

    Response time analysis for prioritized dag task with mutually exclusive vertices

    Ran Bi, Qingqiang He, Jinghao Sun, Zhenyu Sun, Zhishan Guo, Nan Guan, and Guozhen Tan. Response time analysis for prioritized dag task with mutually exclusive vertices. In 2022 IEEE Real-Time Systems Symposium (RTSS), pages 460–473. IEEE, 2022. 30

    2022

  8. [8]

    Brandenburg

    Tobias Blaß, Daniel Casini, Sergey Bozhko, and Bj¨ orn B. Brandenburg. A ROS 2 response- time analysis exploiting starvation freedom and execution-time variance. In42nd IEEE Real- Time Systems Symposium, RTSS 2021, Dortmund, Germany, December 7-10, 2021, pages 41–53. IEEE, 2021

    2021

  9. [9]

    Brandenburg

    Tobias Blaß, Arne Hamann, Ralph Lange, Dirk Ziegenbein, and Bj¨ orn B. Brandenburg. Automatic latency management for ROS 2: Benefits, challenges, and open problems. In 27th IEEE Real-Time and Embedded Technology and Applications Symposium, RTAS 2021, Nashville, TN, USA, May 18-21, 2021, pages 264–277. IEEE, 2021

    2021

  10. [10]

    End to End Learning for Self-Driving Cars

    Mariusz Bojarski, Davide Del Testa, Daniel Dworakowski, Bernhard Firner, Beat Flepp, Prasoon Goyal, Lawrence D Jackel, Mathew Monfort, Urs Muller, Jiakai Zhang, et al. End to end learning for self-driving cars.arXiv preprint arXiv:1604.07316, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  11. [11]

    Adaptive neural net- works for efficient inference

    Tolga Bolukbasi, Joseph Wang, Ofer Dekel, and Venkatesh Saligrama. Adaptive neural net- works for efficient inference. InProceedings of ACM International Conference on Machine Learning, pages 527–536, New York, NY, USA, 2017. ACM

    2017

  12. [12]

    Fully automatic and real-time microrobot detection and tracking based on ultrasound imaging using deep learning

    Karim Botros, Mohammad Alkhatib, David Folio, and Antoine Ferreira. Fully automatic and real-time microrobot detection and tracking based on ultrasound imaging using deep learning. In2022 International Conference on Robotics and Automation (ICRA), pages 9763–9768. IEEE, 2022

    2022

  13. [13]

    Alan Burns and Robert I. Davis. A survey of research into mixed criticality systems.ACM Computing Surveys, 50(6), 2017

    2017

  14. [14]

    Comparative evaluation of the robustness of dag scheduling heuristics.Grid Computing: Achievements and Prospects, pages 73–84, 2008

    Louis-Claude Canon, Emmanuel Jeannot, Rizos Sakellariou, and Wei Zheng. Comparative evaluation of the robustness of dag scheduling heuristics.Grid Computing: Achievements and Prospects, pages 73–84, 2008

    2008

  15. [15]

    Multitask learning.Machine learning, 28(1):41–75, 1997

    Rich Caruana. Multitask learning.Machine learning, 28(1):41–75, 1997

    1997

  16. [16]

    Impact analysis of inference time attack of perception sensors on autonomous vehicles.arXiv preprint arXiv:2505.03850, 2025

    Hanlin Chen, Simin Chen, Wenyu Li, Wei Yang, and Yiheng Feng. Impact analysis of inference time attack of perception sensors on autonomous vehicles.arXiv preprint arXiv:2505.03850, 2025

    work page arXiv 2025

  17. [17]

    Precise response time analysis for multiple dag tasks with intra-task priority assignment

    Nan Chen, Shuai Zhao, Ian Gray, Alan Burns, Siyuan Ji, and Wanli Chang. Precise response time analysis for multiple dag tasks with intra-task priority assignment. In2023 IEEE 29th Real-Time and Embedded Technology and Applications Symposium (RTAS), pages 174–184. IEEE, 2023

    2023

  18. [18]

    Nmtsloth: understanding and testing efficiency degradation of neural machine translation systems

    Simin Chen, Cong Liu, Mirazul Haque, Zihe Song, and Wei Yang. Nmtsloth: understanding and testing efficiency degradation of neural machine translation systems. InProceedings of the 30th ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering, pages 1148–1160, 2022

    2022

  19. [19]

    Your compiler is back- dooring your model: Understanding and exploiting compilation inconsistency vulnerabilities in deep learning compilers.arXiv preprint arXiv:2509.11173, 2025

    Simin Chen, Jinjun Peng, Yixin He, Junfeng Yang, and Baishakhi Ray. Your compiler is back- dooring your model: Understanding and exploiting compilation inconsistency vulnerabilities in deep learning compilers.arXiv preprint arXiv:2509.11173, 2025

    work page arXiv 2025

  20. [20]

    Dycodeeval: Dynamic benchmarking of reasoning capabilities in code large language models under data contamination

    Simin Chen, Pranav Pusarla, and Baishakhi Ray. Dycodeeval: Dynamic benchmarking of reasoning capabilities in code large language models under data contamination. InForty- second International Conference on Machine Learning. 31

  21. [21]

    Nicgslowdown: Evaluating the efficiency robustness of neural image caption generation models

    Simin Chen, Zihe Song, Mirazul Haque, Cong Liu, and Wei Yang. Nicgslowdown: Evaluating the efficiency robustness of neural image caption generation models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15365–15374, 2022

    2022

  22. [22]

    Dycl: Dynamic neural network compilation via program rewriting and graph optimization

    Simin Chen, Shiyi Wei, Cong Liu, and Wei Yang. Dycl: Dynamic neural network compilation via program rewriting and graph optimization. InProceedings of the 32nd ACM SIGSOFT International Symposium on Software Testing and Analysis, pages 614–626, 2023

    2023

  23. [23]

    Tan, and Haizhou Li

    Yiming Chen, Simin Chen, Zexin Li, Wei Yang, Cong Liu, Robby T. Tan, and Haizhou Li. Dynamic transformers provide a false sense of efficiency. In Anna Rogers, Jordan L. Boyd-Graber, and Naoaki Okazaki, editors,Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2023, Toronto, Canada, July 9-...

    2023

  24. [24]

    Beyond single-audio: Advancing multi-audio processing in audio large language models.arXiv preprint arXiv:2409.18680, 2024

    Yiming Chen, Xianghu Yue, Xiaoxue Gao, Chen Zhang, Luis Fernando D’Haro, Robby T Tan, and Haizhou Li. Beyond single-audio: Advancing multi-audio processing in audio large language models.arXiv preprint arXiv:2409.18680, 2024

    work page arXiv 2024

  25. [25]

    VoiceBench: Benchmarking LLM-Based Voice Assistants

    Yiming Chen, Xianghu Yue, Chen Zhang, Xiaoxue Gao, Robby T Tan, and Haizhou Li. Voicebench: Benchmarking llm-based voice assistants.arXiv preprint arXiv:2410.17196, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  26. [26]

    Unveiling the achilles’ heel of nlg evaluators: A unified adversarial framework driven by large language models.arXiv preprint arXiv:2405.14646, 2024

    Yiming Chen, Chen Zhang, Danqing Luo, Luis Fernando D’Haro, Robby T Tan, and Haizhou Li. Unveiling the achilles’ heel of nlg evaluators: A unified adversarial framework driven by large language models.arXiv preprint arXiv:2405.14646, 2024

    work page arXiv 2024

  27. [27]

    Revisiting self-training for few-shot learning of language model.arXiv preprint arXiv:2110.01256, 2021

    Yiming Chen, Yan Zhang, Chen Zhang, Grandee Lee, Ran Cheng, and Haizhou Li. Revisiting self-training for few-shot learning of language model.arXiv preprint arXiv:2110.01256, 2021

    work page arXiv 2021

  28. [28]

    Picas: New design of priority-driven chain-aware scheduling for ROS2

    Hyunjong Choi, Yecheng Xiang, and Hyoseung Kim. Picas: New design of priority-driven chain-aware scheduling for ROS2. In27th IEEE Real-Time and Embedded Technology and Applications Symposium, RTAS 2021, Nashville, TN, USA, May 18-21, 2021, pages 251–263. IEEE, 2021

    2021

  29. [29]

    Compressing neural networks using the variational information bottleneck

    Bin Dai, Chen Zhu, Baining Guo, and David Wipf. Compressing neural networks using the variational information bottleneck. InInternational Conference on Machine Learning, pages 1135–1144. PMLR, 2018

    2018

  30. [30]

    Imagenet: A large- scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large- scale hierarchical image database. In2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009

    2009

  31. [31]

    Nestdnn: Resource-aware multi-tenant on-device deep learning for continuous mobile vision

    Biyi Fang, Xiao Zeng, and Mi Zhang. Nestdnn: Resource-aware multi-tenant on-device deep learning for continuous mobile vision. InProceedings of ACM Annual International Conference on Mobile Computing and Networking, pages 115–127, 2018

    2018

  32. [32]

    Llmeffichecker: Understanding and testing efficiency degradation of large language models.ACM Transactions on Software Engineering and Methodology, 33(7):1–38, 2024

    Xiaoning Feng, Xiaohong Han, Simin Chen, and Wei Yang. Llmeffichecker: Understanding and testing efficiency degradation of large language models.ACM Transactions on Software Engineering and Methodology, 33(7):1–38, 2024. 32

    2024

  33. [33]

    Mega evolving graph accelerator

    Chao Gao, Mahbod Afarin, Shafiur Rahman, Nael Abu-Ghazaleh, and Rajiv Gupta. Mega evolving graph accelerator. InProceedings of the 56th Annual IEEE/ACM International Symposium on Microarchitecture, MICRO ’23, page 310–323, New York, NY, USA, 2023. Association for Computing Machinery

    2023

  34. [34]

    Dlora: Distributed parameter-efficient fine-tuning solution for large language model.arXiv preprint arXiv:2404.05182, 2024

    Chao Gao and Sai Qian Zhang. Dlora: Distributed parameter-efficient fine-tuning solution for large language model.arXiv preprint arXiv:2404.05182, 2024

    work page arXiv 2024

  35. [35]

    Are we ready for autonomous driving? the kitti vision benchmark suite

    Andreas Geiger, Philip Lenz, and Raquel Urtasun. Are we ready for autonomous driving? the kitti vision benchmark suite. In2012 IEEE conference on computer vision and pattern recognition, pages 3354–3361. IEEE, 2012

    2012

  36. [36]

    Enabling low-cost secure computing on untrusted in-memory architectures

    Sahar Ghoflsaz Ghinani, Jingyao Zhang, and Elaheh Sadredini. Enabling low-cost secure computing on untrusted in-memory architectures

  37. [37]

    D3: a dynamic deadline-driven approach for building autonomous vehicles

    Ionel Gog, Sukrit Kalra, Peter Schafhalter, Joseph E Gonzalez, and Ion Stoica. D3: a dynamic deadline-driven approach for building autonomous vehicles. InProceedings of the Seventeenth European Conference on Computer Systems, pages 453–471, 2022

    2022

  38. [38]

    Pylot: A modular platform for exploring latency-accuracy tradeoffs in autonomous vehicles

    Ionel Gog, Sukrit Kalra, Peter Schafhalter, Matthew A Wright, Joseph E Gonzalez, and Ion Stoica. Pylot: A modular platform for exploring latency-accuracy tradeoffs in autonomous vehicles. In2021 IEEE International Conference on Robotics and Automation (ICRA), pages 8806–8813. IEEE, 2021

    2021

  39. [39]

    Ronald L. Graham. Bounds on multiprocessing timing anomalies.SIAM journal on Applied Mathematics, 17(2):416–429, 1969

    1969

  40. [40]

    Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey

    Zeyu Han, Chao Gao, Jinyang Liu, Sai Qian Zhang, et al. Parameter-efficient fine-tuning for large models: A comprehensive survey.arXiv preprint arXiv:2403.14608, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  41. [41]

    Bounding the response time of dag tasks using long paths

    Qingqiang He, Nan Guan, Mingsong Lv, Xu Jiang, and Wanli Chang. Bounding the response time of dag tasks using long paths. In2022 IEEE Real-Time Systems Symposium (RTSS), pages 474–486. IEEE, 2022

    2022

  42. [42]

    Robust electric vehicle balancing of autonomous mobility-on-demand system: A multi-agent reinforcement learning approach.arXiv preprint arXiv:2307.16228, 2023

    Sihong He, Shuo Han, and Fei Miao. Robust electric vehicle balancing of autonomous mobility-on-demand system: A multi-agent reinforcement learning approach.arXiv preprint arXiv:2307.16228, 2023

    work page arXiv 2023

  43. [43]

    Robust multi-agent reinforcement learning with state uncertainty.Transactions on Machine Learning Research, 2023

    Sihong He, Songyang Han, Sanbao Su, Shuo Han, Shaofeng Zou, and Fei Miao. Robust multi-agent reinforcement learning with state uncertainty.Transactions on Machine Learning Research, 2023

    2023

  44. [44]

    A robust and constrained multi-agent reinforcement learning framework for electric vehicle amod systems.arXiv preprint arXiv:2209.08230, 2022

    Sihong He, Yue Wang, Shuo Han, Shaofeng Zou, and Fei Miao. A robust and constrained multi-agent reinforcement learning framework for electric vehicle amod systems.arXiv preprint arXiv:2209.08230, 2022

    work page arXiv 2022

  45. [45]

    On-device deep multi-task inference via multi-task zipping.IEEE Transactions on Mobile Computing, pages 1–1, 2021

    Xiaoxi He, Xu Wang, Zimu Zhou, Jiahang Wu, Zheng Yang, and Lothar Thiele. On-device deep multi-task inference via multi-task zipping.IEEE Transactions on Mobile Computing, pages 1–1, 2021

    2021

  46. [46]

    Multi-task zipping via layer-wise neuron sharing

    Xiaoxi He, Zimu Zhou, and Lothar Thiele. Multi-task zipping via layer-wise neuron sharing. InAdvances in Neural Information Processing Systems, pages 6019–6029, 2018. 33

    2018

  47. [47]

    Ros2 real-time classification and detection.https://github.com/ NVIDIA-AI-IOT/ros2_torch_trt, 2020

    NVIDIA AI IOT. Ros2 real-time classification and detection.https://github.com/ NVIDIA-AI-IOT/ros2_torch_trt, 2020

    2020

  48. [48]

    Pccl: Energy-efficient llm training with power-aware collective communication

    Ziyang Jia, Laxmi N Bhuyan, and Daniel Wong. Pccl: Energy-efficient llm training with power-aware collective communication. In2024 IEEE 42nd International Conference on Computer Design (ICCD), pages 84–91. IEEE, 2024

    2024

  49. [49]

    Fine-grained warm water cooling for improving datacenter economy

    Weixiang Jiang, Ziyang Jia, Sirui Feng, Fangming Liu, and Hai Jin. Fine-grained warm water cooling for improving datacenter economy. InProceedings of the 46th International Symposium on Computer Architecture, pages 474–486, 2019

    2019

  50. [50]

    Real-time scheduling of parallel tasks with tight deadlines.Journal of Systems Architecture, 108:101742, 2020

    Xu Jiang, Nan Guan, Xiang Long, Yue Tang, and Qingqiang He. Real-time scheduling of parallel tasks with tight deadlines.Journal of Systems Architecture, 108:101742, 2020

    2020

  51. [51]

    Semi-federated scheduling of parallel real- time tasks on multiprocessors

    Xu Jiang, Nan Guan, Xiang Long, and Wang Yi. Semi-federated scheduling of parallel real- time tasks on multiprocessors. In2017 IEEE Real-Time Systems Symposium (RTSS), pages 80–91. IEEE, 2017

    2017

  52. [52]

    Real-time scheduling and analysis of processing chains on multi-threaded executor in ROS 2

    Xu Jiang, Dong Ji, Nan Guan, Ruoxiang Li, Yue Tang, and Yi Wang. Real-time scheduling and analysis of processing chains on multi-threaded executor in ROS 2. InIEEE Real-Time Systems Symposium, RTSS 2022, Houston, TX, USA, December 5-8, 2022, pages 27–39. IEEE, 2022

    2022

  53. [53]

    Rt-mot: Confidence-aware real-time scheduling framework for multi-object tracking tasks

    Donghwa Kang, Seunghoon Lee, Hoon Sung Chwa, Seung-Hwan Bae, Chang Mook Kang, Jinkyu Lee, and Hyeongboo Baek. Rt-mot: Confidence-aware real-time scheduling framework for multi-object tracking tasks. In2022 IEEE Real-Time Systems Symposium (RTSS), pages 318–330. IEEE, 2022

    2022

  54. [54]

    Rt-swap: Addressing gpu memory bottlenecks for real-time multi-dnn inference

    Woosung Kang, Jinkyu Lee, Youngmoon Lee, Sangeun Oh, Kilho Lee, and Hoon Sung Chwa. Rt-swap: Addressing gpu memory bottlenecks for real-time multi-dnn inference. In2024 IEEE 30th Real-Time and Embedded Technology and Applications Symposium (RTAS), pages 373–385. IEEE Computer Society, 2024

    2024

  55. [55]

    Lalarand: Flexible layer-by-layer cpu/gpu scheduling for real-time dnn tasks

    Woosung Kang, Kilho Lee, Jinkyu Lee, Insik Shin, and Hoon Sung Chwa. Lalarand: Flexible layer-by-layer cpu/gpu scheduling for real-time dnn tasks. In2021 IEEE Real-Time Systems Symposium (RTSS), pages 329–341. IEEE, 2021

    2021

  56. [56]

    Xrbench: An extended reality (xr) machine learning benchmark suite for the metaverse.arXiv preprint arXiv:2211.08675, 2022

    Hyoukjun Kwon, Krishnakumar Nair, Jamin Seo, Jason Yik, Debabrata Mohapatra, Dongyuan Zhan, Jinook Song, Peter Capak, Peizhao Zhang, Peter Vajda, et al. Xrbench: An extended reality (xr) machine learning benchmark suite for the metaverse.arXiv preprint arXiv:2211.08675, 2022

    work page arXiv 2022

  57. [57]

    Towards safe, realistic testbed for robotic systems with human interaction

    Bhoram Lee, Jonathan Brookshire, Rhys Yahata, and Supun Samarasekera. Towards safe, realistic testbed for robotic systems with human interaction. In2022 International Conference on Robotics and Automation (ICRA), pages 11280–11287. IEEE, 2022

    2022

  58. [58]

    Fast and scalable in-memory deep multitask learning via neural weight virtualization

    Seulki Lee and Shahriar Nirjon. Fast and scalable in-memory deep multitask learning via neural weight virtualization. InProceedings of ACM Annual International Conference on Mobile Systems, Applications, and Services, pages 175–190, 2020

    2020

  59. [59]

    Outstanding paper award: Analysis of global edf for parallel tasks

    Jing Li, Kunal Agrawal, Chenyang Lu, and Christopher Gill. Outstanding paper award: Analysis of global edf for parallel tasks. In2013 25th Euromicro Conference on Real-Time Systems, pages 3–13. IEEE, 2013. 34

    2013

  60. [60]

    Analysis of federated and global scheduling for parallel real-time tasks

    Jing Li, Jian Jia Chen, Kunal Agrawal, Chenyang Lu, Chris Gill, and Abusayeed Saiful- lah. Analysis of federated and global scheduling for parallel real-time tasks. In2014 26th Euromicro Conference on Real-Time Systems, pages 85–96. IEEE, 2014

    2014

  61. [61]

    Worst- case time disparity analysis of message synchronization in ROS

    Ruoxiang Li, Nan Guan, Xu Jiang, Zhishan Guo, Zheng Dong, and Mingsong Lv. Worst- case time disparity analysis of message synchronization in ROS. InIEEE Real-Time Systems Symposium, RTSS 2022, Houston, TX, USA, December 5-8, 2022, pages 40–52. IEEE, 2022

    2022

  62. [62]

    Mace: A hybrid llm serving system with colocated slo-aware continuous retraining alignment.arXiv preprint arXiv:2510.03283, 2025

    Yufei Li, Yu Fu, Yue Dong, and Cong Liu. Mace: A hybrid llm serving system with colocated slo-aware continuous retraining alignment.arXiv preprint arXiv:2510.03283, 2025

    work page arXiv 2025

  63. [63]

    White-box multi-objective adversarial attack on dialogue generation

    Yufei Li, Zexin Li, Yingfan Gao, and Cong Liu. White-box multi-objective adversarial attack on dialogue generation. In Anna Rogers, Jordan L. Boyd-Graber, and Naoaki Okazaki, edi- tors,Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2023, Toronto, Canada, July 9-14, 2023, pages 1778–1792...

    2023

  64. [64]

    Rt-lm: Uncertainty-aware resource management for real-time inference of language models

    Yufei Li, Zexin Li, Wei Yang, and Cong Liu. Rt-lm: Uncertainty-aware resource management for real-time inference of language models. In2023 IEEE Real-Time Systems Symposium (RTSS), pages 158–171. IEEE, 2023

    2023

  65. [65]

    Lemix: Unified scheduling for llm training and inference on multi-gpu systems.arXiv preprint arXiv:2507.21276, 2025

    Yufei Li, Zexin Li, Yinglun Zhu, and Cong Liu. Lemix: Unified scheduling for llm training and inference on multi-gpu systems.arXiv preprint arXiv:2507.21276, 2025

    work page arXiv 2025

  66. [66]

    Genie: Smart ros-based caching for connected autonomous robots.arXiv preprint arXiv:2402.19410, 2024

    Zexin Li, Soroush Bateni, and Cong Liu. Genie: Smart ros-based caching for connected autonomous robots.arXiv preprint arXiv:2402.19410, 2024

    work page arXiv 2024

  67. [67]

    Mi- monet: Multi-input multi-output on-device deep learning.arXiv preprint arXiv:2307.11962, 2023

    Zexin Li, Xiaoxi He, Yufei Li, Shahab Nikkhoo, Wei Yang, Lothar Thiele, and Cong Liu. Mi- monet: Multi-input multi-output on-device deep learning.arXiv preprint arXiv:2307.11962, 2023

    work page arXiv 2023

  68. [68]

    RED: A systematic real-time scheduling approach for robotic environmental dynamics

    Zexin Li, Tao Ren, Xiaoxi He, and Cong Liu. RED: A systematic real-time scheduling approach for robotic environmental dynamics. In2023 IEEE Real-Time Systems Symposium (RTSS), pages 210–223. IEEE, 2023

    2023

  69. [69]

    r 3: On-device real-time deep reinforcement learning for autonomous robotics

    Zexin Li, Aritra Samanta, Yufei Li, Andrea Soltoggio, Hyoseung Kim, and Cong Liu. r 3: On-device real-time deep reinforcement learning for autonomous robotics. In2023 IEEE Real-Time Systems Symposium (RTSS), pages 131–144. IEEE, 2023

    2023

  70. [70]

    Efficient algorithms for task mapping on heterogeneous cpu/gpu platforms for fast completion time.Journal of Systems Architecture, 114:101936, 2021

    Zexin Li, Yuqun Zhang, Ao Ding, Husheng Zhou, and Cong Liu. Efficient algorithms for task mapping on heterogeneous cpu/gpu platforms for fast completion time.Journal of Systems Architecture, 114:101936, 2021

    2021

  71. [71]

    Prentice Hall PTR, 2000

    Jane WSW Liu.Real-time systems. Prentice Hall PTR, 2000

    2000

  72. [72]

    Fisheyemultinet: Real-time multi-task learning architecture for surround-view automated parking system.arXiv preprint arXiv:1912.11066, 2019

    Pullarao Maddu, Wayne Doherty, Ganesh Sistu, Isabelle Leang, Michal Uricar, Sumanth Chennupati, Hazem Rashed, Jonathan Horgan, Ciaran Hughes, and Senthil Yogamani. Fisheyemultinet: Real-time multi-task learning architecture for surround-view automated parking system.arXiv preprint arXiv:1912.11066, 2019

    work page arXiv 1912

  73. [73]

    Deepeye: Resource efficient local execution of multiple deep vision models us- ing wearable commodity hardware

    Akhil Mathur, Nicholas D Lane, Sourav Bhattacharya, Aidan Boran, Claudio Forlivesi, and Fahim Kawsar. Deepeye: Resource efficient local execution of multiple deep vision models us- ing wearable commodity hardware. InProceedings of ACM Annual International Conference on Mobile Systems, Applications, and Services, pages 68–81, 2017. 35

    2017

  74. [74]

    Response-time analysis of conditional dag tasks in multiprocessor sys- tems

    Alessandra Melani, Marko Bertogna, Vincenzo Bonifaci, Alberto Marchetti-Spaccamela, and Giorgio C Buttazzo. Response-time analysis of conditional dag tasks in multiprocessor sys- tems. In2015 27th Euromicro Conference on Real-Time Systems, pages 211–221. IEEE, 2015

    2015

  75. [75]

    Pimbot: Policy and incentive manipulation for multi-robot reinforcement learning in social dilemmas

    Shahab Nikkhoo, Zexin Li, Aritra Samanta, Yufei Li, and Cong Liu. Pimbot: Policy and incentive manipulation for multi-robot reinforcement learning in social dilemmas. In2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 5630–

  76. [76]

    Duckiebot (db-j).https://get.duckietown.com/products/duckiebot-db21, 2022

    NVIDIA. Duckiebot (db-j).https://get.duckietown.com/products/duckiebot-db21, 2022

    2022

  77. [77]

    Sparkfun jetbot ai kit.https://www.sparkfun.com/products/18486, 2022

    NVIDIA. Sparkfun jetbot ai kit.https://www.sparkfun.com/products/18486, 2022

    2022

  78. [78]

    Waveshare jetbot ai kit.https://www.amazon.com/ Waveshare-JetBot-AI-Kit-Accessories/dp/B07V8JL4TF/, 2022

    NVIDIA. Waveshare jetbot ai kit.https://www.amazon.com/ Waveshare-JetBot-AI-Kit-Accessories/dp/B07V8JL4TF/, 2022

    2022

  79. [79]

    Virtual to Real Reinforcement Learning for Autonomous Driving

    Xinlei Pan, Yurong You, Ziyan Wang, and Cewu Lu. Virtual to real reinforcement learning for autonomous driving.arXiv preprint arXiv:1704.03952, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  80. [80]

    A framework for real-time service-oriented architecture

    Mark Panahi, Weiran Nie, and Kwei-Jay Lin. A framework for real-time service-oriented architecture. In2009 IEEE Conference on Commerce and Enterprise Computing, pages 460–

Showing first 80 references.