kth.sePublications
1 - 2 of 2
rss atomLink to result list
Permanent link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
  • Presentation: 2025-02-21 10:00 D37, Stockholm
    Duan, Zhihao
    KTH, School of Engineering Sciences (SCI), Engineering Mechanics.
    Motor Unit Property Estimation and Clustering in Individuals with Spinal Cord Injury2025Licentiate thesis, comprehensive summary (Other academic)
    Abstract [en]

    Spinal cord injury (SCI) often results in various alternations in muscle functions, including disrupted muscle activation patterns and impaired coordination. These changes can be quantitatively assessed using traditional electromyography. However, in vivo assessment of electrophysiological parameters of motor units (MUs) remains challenging, limiting our understanding of neuromuscular mechanisms underlying these alternations. High-density electromyography (HD-EMG) with decomposition algorithms offers new opportunities to investigate MU properties and firing behavior in vivo. Despite these advances, MU electrophysiological parameters – critical for MU recruitment, firing patterns, and MU synergies after SCI have not been thoroughly investigated. This thesis presented two studies addressing the gaps, focusing on MU parameter alterations and clustering post-SCI.       

    In the first study, we proposed an integrated approach combining HD-EMG and motor neuron modeling to estimate key MU electrophysiological parameters: soma size and inert period. These parameters are crucial for understanding MU recruitment and firing patterns. HD-EMG and ankle torque were collected simultaneously on tibialis anterior, soleus (SOL), and gastrocnemius medialis (GM) muscles during submaximal isometric dorsiflexion and plantar flexion tasks in both participants with SCI and able-bodied subjects. Comparisons between groups revealed a significantly longer inert period in the tibialis anterior muscle among individuals with SCI, suggesting delayed MU recovery times required for the MU to be re-excited,  potentially leading to decreased firing rates. However, the limited number of decomposed MUs, particularly at higher contraction levels, restricted the ability to fully capture the differences between groups and across the SOL and GM muscles. Our analysis further demonstrated that the proposed approach could reliably estimate MU electrophysiological parameters in vivo, offering valuable insights for personalized assessment and monitoring of MU properties in clinical populations.

    In the second study, we examined MU synergies and clustering in the synergetic ankle plantarflexors SOL and GM muscles during 20% and 50% maximal voluntary contraction and explored how these patterns were altered following SCI. To evaluate the shared neural drive, we calculated the coherence between the MUs between the SOL and GM muscles. Factor analysis was employed to extract MU modes for each muscle and the decomposed MUs were categorized into distinct functional groups based on their correlations with each mode. The results demonstrated significant coherence between the SOL and GM muscles in both groups, indicating a strong shared neural drive that facilitates their coordinated function. In the SCI group, the results showed significantly higher coherence in the delta frequency band at 50% maximal voluntary contraction compared to the control group, suggesting a disrupted muscle coordination after SCI. The clustering results showed a significantly reduced proportion of the shared cluster within GM muscle in the SCI group at 20% maximal voluntary contraction, indicating a disrupted MU clustering, potentially affecting motor coordination.  As the contraction level increased, the control group exhibited a decrease in the proportion of the shared cluster and an increase in the proportion of the self cluster. In contrast, no significant changes were observed in the SCI group. 

    Together, these studies presented novel approaches to estimating MU properties and clustering in vivo, offering valuable insight into the MU electrophysiological parameters and MU synergies adaptation after SCI. These findings could inform the development of advanced rehabilitation strategies and enhance intervention outcomes. 

    Download (pdf)
    kappa
  • Presentation: 2025-02-21 15:00 https://kth-se.zoom.us/j/66674834407, Stockholm
    Ruggeri, Franco
    KTH, School of Electrical Engineering and Computer Science (EECS), Intelligent systems, Decision and Control Systems (Automatic Control).
    Explainable Reinforcement Learning for Mobile Network Optimization2025Licentiate thesis, comprehensive summary (Other academic)
    Abstract [en]

    The growing complexity of mobile networks has driven the need for automated optimization approaches, with Reinforcement Learning (RL) emerging as a promising data-driven technique for controlling network parameters. However, RL systems often operate as black boxes, lacking the interpretability and transparency required by Mobile Network Operators (MNOs) and Artificial Intelligence (AI) engineers to trust, monitor, and refine their behavior. This lack poses significant challenges, particularly in the telecommunications domain, where ensuring alignment with operational objectives and maintaining reliable network performance is critical. Consequently, there is a pressing need for explainability methods that make RL decisions more interpretable and understandable for stakeholders.

    This thesis investigates the emerging field of Explainable Reinforcement Learning (XRL), specifically focusing on its application to mobile network optimization. In the context of single-agent RL, we evaluate two state-of-the-art XRL techniques in the Remote Electrical Tilt (RET) optimization problem, where the tilt of each antenna needs to be controlled to optimize network coverage and capacity. These methods address two distinct interpretability challenges in RL: (i) understanding the state-action mapping determined by an RL policy and (ii) explaining the long-term goal of an RL agent. These evaluations highlight the potential and limitations of existing XRL methods when applied to a simulated mobile network.

    To address a significant gap in the literature on single-agent XRL, we devise a novel algorithm, Temporal Policy Decomposition (TPD), which explains RL actions by predicting their outcomes in upcoming time steps. This method provides a clear view of an agent's anticipated behavior starting from a given environment state by generating insights for individual time steps. These time-aware explanations offer a comprehensive understanding of the decision-making process that accounts for the sequential nature of RL.

    We then focus on multi-agent systems and develop a rollout-based algorithm to estimate Local Shapley Values (LSVs), quantifying individual agent contributions in specific states. This method reliably identifies agent contributions even in scenarios involving undertrained or suboptimal agents, making it a valuable tool for monitoring and diagnosing cooperative multi-agent systems.

    These contributions represent a step toward a holistic explainability framework for RL in mobile networks, combining single-agent and multi-agent perspectives. By addressing core interpretability challenges, this research equips MNOs and AI engineers with practical techniques to trust, monitor, debug, and refine RL models. Furthermore, it helps ensure readiness for potential regulatory requirements, contributing to the broader goal of advancing trustworthy AI in telecommunications.

    Download full text (pdf)
    fulltext