---
_id: '66147'
abstract:
- lang: eng
  text: |-
    <jats:title>Abstract</jats:title>
              <jats:p>Reinforcement Learning (RL) has emerged as a promising tool for decision-making in various applications, particularly in uncertain environments. While its adoption in embedded systems—especially hard real-time systems—faces challenges due to stringent timing constraints, integrating shielding mechanisms may offer a pathway for RL to optimize its scheduling decisions, preserving worst-case timing guarantees. This position paper shows a use case where RL selects compliant execution versions for fault-tolerant real-time systems while minimizing the system utilization in runtime. Furthermore, we discuss possible directions for further exploring RL’s role in real-time systems for improved adaptability.</jats:p>
author:
- first_name: Junjie
  full_name: Shi, Junjie
  last_name: Shi
- first_name: Kuan-Hsun
  full_name: Chen, Kuan-Hsun
  last_name: Chen
citation:
  ama: Shi J, Chen K-H. Shielded reinforcement learning for fault-tolerant scheduling
    in real-time systems. <i>Real-Time Systems</i>. 2025;61(2):306-310. doi:<a href="https://doi.org/10.1007/s11241-025-09441-z">10.1007/s11241-025-09441-z</a>
  apa: Shi, J., &#38; Chen, K.-H. (2025). Shielded reinforcement learning for fault-tolerant
    scheduling in real-time systems. <i>Real-Time Systems</i>, <i>61</i>(2), 306–310.
    <a href="https://doi.org/10.1007/s11241-025-09441-z">https://doi.org/10.1007/s11241-025-09441-z</a>
  bibtex: '@article{Shi_Chen_2025, title={Shielded reinforcement learning for fault-tolerant
    scheduling in real-time systems}, volume={61}, DOI={<a href="https://doi.org/10.1007/s11241-025-09441-z">10.1007/s11241-025-09441-z</a>},
    number={2}, journal={Real-Time Systems}, publisher={Springer Science and Business
    Media LLC}, author={Shi, Junjie and Chen, Kuan-Hsun}, year={2025}, pages={306–310}
    }'
  chicago: 'Shi, Junjie, and Kuan-Hsun Chen. “Shielded Reinforcement Learning for
    Fault-Tolerant Scheduling in Real-Time Systems.” <i>Real-Time Systems</i> 61,
    no. 2 (2025): 306–10. <a href="https://doi.org/10.1007/s11241-025-09441-z">https://doi.org/10.1007/s11241-025-09441-z</a>.'
  ieee: 'J. Shi and K.-H. Chen, “Shielded reinforcement learning for fault-tolerant
    scheduling in real-time systems,” <i>Real-Time Systems</i>, vol. 61, no. 2, pp.
    306–310, 2025, doi: <a href="https://doi.org/10.1007/s11241-025-09441-z">10.1007/s11241-025-09441-z</a>.'
  mla: Shi, Junjie, and Kuan-Hsun Chen. “Shielded Reinforcement Learning for Fault-Tolerant
    Scheduling in Real-Time Systems.” <i>Real-Time Systems</i>, vol. 61, no. 2, Springer
    Science and Business Media LLC, 2025, pp. 306–10, doi:<a href="https://doi.org/10.1007/s11241-025-09441-z">10.1007/s11241-025-09441-z</a>.
  short: J. Shi, K.-H. Chen, Real-Time Systems 61 (2025) 306–310.
date_created: 2026-07-03T21:07:18Z
date_updated: 2026-07-05T14:48:45Z
doi: 10.1007/s11241-025-09441-z
intvolume: '        61'
issue: '2'
language:
- iso: eng
page: 306-310
publication: Real-Time Systems
publication_identifier:
  issn:
  - 0922-6443
  - 1573-1383
publication_status: published
publisher: Springer Science and Business Media LLC
status: public
title: Shielded reinforcement learning for fault-tolerant scheduling in real-time
  systems
type: journal_article
user_id: '128464'
volume: 61
year: '2025'
...
---
_id: '63059'
abstract:
- lang: eng
  text: "<jats:title>Abstract</jats:title><jats:p>While high accuracy is of paramount
    importance for deep learning (DL) inference, serving inference requests on time
    is equally critical but has not been carefully studied especially when the request
    has to be served over a dynamic wireless network at the edge. In this paper, we
    propose Jellyfish—a novel edge DL inference serving system that achieves soft
    guarantees for end-to-end inference latency service-level objectives (SLO). Jellyfish
    handles the network variability by utilizing both data and deep neural network
    (DNN) adaptation to conduct tradeoffs between accuracy and latency. Jellyfish
    features a new design that enables collective adaptation policies where the decisions
    for data and DNN adaptations are aligned and coordinated among multiple users
    with varying network conditions. We propose efficient algorithms to continuously
    map users and adapt DNNs at runtime, so that we fulfill latency SLOs while maximizing
    the overall inference accuracy. We further investigate <jats:italic>dynamic</jats:italic>
    DNNs, i.e., DNNs that encompass multiple architecture variants, and demonstrate
    their potential benefit through preliminary experiments. Our experiments based
    on a prototype implementation and real-world WiFi and LTE network traces show
    that Jellyfish can meet latency SLOs at around the 99th percentile while maintaining
    high accuracy.\r\n</jats:p>"
author:
- first_name: Vinod
  full_name: Nigade, Vinod
  last_name: Nigade
- first_name: Pablo
  full_name: Bauszat, Pablo
  last_name: Bauszat
- first_name: Henri
  full_name: Bal, Henri
  last_name: Bal
- first_name: Lin
  full_name: Wang, Lin
  id: '102868'
  last_name: Wang
  orcid: 0000-0001-7181-6128
citation:
  ama: Nigade V, Bauszat P, Bal H, Wang L. Inference serving with end-to-end latency
    SLOs over dynamic edge networks. <i>Real-Time Systems</i>. 2024;60(2):239-290.
    doi:<a href="https://doi.org/10.1007/s11241-024-09418-4">10.1007/s11241-024-09418-4</a>
  apa: Nigade, V., Bauszat, P., Bal, H., &#38; Wang, L. (2024). Inference serving
    with end-to-end latency SLOs over dynamic edge networks. <i>Real-Time Systems</i>,
    <i>60</i>(2), 239–290. <a href="https://doi.org/10.1007/s11241-024-09418-4">https://doi.org/10.1007/s11241-024-09418-4</a>
  bibtex: '@article{Nigade_Bauszat_Bal_Wang_2024, title={Inference serving with end-to-end
    latency SLOs over dynamic edge networks}, volume={60}, DOI={<a href="https://doi.org/10.1007/s11241-024-09418-4">10.1007/s11241-024-09418-4</a>},
    number={2}, journal={Real-Time Systems}, publisher={Springer Science and Business
    Media LLC}, author={Nigade, Vinod and Bauszat, Pablo and Bal, Henri and Wang,
    Lin}, year={2024}, pages={239–290} }'
  chicago: 'Nigade, Vinod, Pablo Bauszat, Henri Bal, and Lin Wang. “Inference Serving
    with End-to-End Latency SLOs over Dynamic Edge Networks.” <i>Real-Time Systems</i>
    60, no. 2 (2024): 239–90. <a href="https://doi.org/10.1007/s11241-024-09418-4">https://doi.org/10.1007/s11241-024-09418-4</a>.'
  ieee: 'V. Nigade, P. Bauszat, H. Bal, and L. Wang, “Inference serving with end-to-end
    latency SLOs over dynamic edge networks,” <i>Real-Time Systems</i>, vol. 60, no.
    2, pp. 239–290, 2024, doi: <a href="https://doi.org/10.1007/s11241-024-09418-4">10.1007/s11241-024-09418-4</a>.'
  mla: Nigade, Vinod, et al. “Inference Serving with End-to-End Latency SLOs over
    Dynamic Edge Networks.” <i>Real-Time Systems</i>, vol. 60, no. 2, Springer Science
    and Business Media LLC, 2024, pp. 239–90, doi:<a href="https://doi.org/10.1007/s11241-024-09418-4">10.1007/s11241-024-09418-4</a>.
  short: V. Nigade, P. Bauszat, H. Bal, L. Wang, Real-Time Systems 60 (2024) 239–290.
date_created: 2025-12-12T08:16:33Z
date_updated: 2025-12-12T08:18:05Z
department:
- _id: '34'
- _id: '7'
- _id: '75'
doi: 10.1007/s11241-024-09418-4
intvolume: '        60'
issue: '2'
language:
- iso: eng
page: 239-290
publication: Real-Time Systems
publication_identifier:
  issn:
  - 0922-6443
  - 1573-1383
publication_status: published
publisher: Springer Science and Business Media LLC
status: public
title: Inference serving with end-to-end latency SLOs over dynamic edge networks
type: journal_article
user_id: '102868'
volume: 60
year: '2024'
...