Hierarchical MARL for Industrial Maintenance
Focus: Hierarchical decision-making, SMDP options, tactical MARL, maintenance scheduling, constrained policies.
Research
Decision-making systems under uncertainty — reinforcement learning, multi-agent coordination, human-in-the-loop control, and industrial maintenance optimization. Peer-reviewed publications, manuscripts in progress, and concept notes are listed separately below.
Peer-reviewed / Published
| Title | Status | Venue | Year | Links |
|---|---|---|---|---|
| Utilizing the Dempster–Shafer Theory for Condition Diagnosis of Continuous Casting Machines | Published | VIVT Journal | TBD | PDF · Demo · Project page |
| Hierarchical decision-making under uncertainty — main research article (Russian) | Accepted | TBD | TBD | eLibrary |
| Proof-of-concept implementation — related publication (Russian) | Accepted | TBD | TBD | eLibrary |
Manuscripts in Progress
| Title | Status | Venue | Year | Links |
|---|---|---|---|---|
| Hierarchical multi-agent reinforcement learning framework for maintenance and repair (MRO) | Working paper | TBD | TBD | Draft / note |
| Marginal Hierarchical Reinforcement Learning for Asymmetric Multi-Agent Systems | Concept note | TBD | 2025 | Draft / note |
Research Directions
Active research tracks — not yet standalone publications.
Human-in-the-Loop Reinforcement Learning
Focus: Interval HITL operator, latency-aware decision loops, operator uncertainty, fallback logic, safety constraints.
Evidential Decision Processes (DST-MDP)
Focus: Dempster–Shafer theory, belief/plausibility functions, DST-MDP, robust decision-making under incomplete information.
Resource-Aware / Marginal HRL
Focus: Allocation of computational or operational resources between agents, leadership migration, marginal utility of control.
Collaborations / Side Research
- Participation in CayleyPy Project — Suspended. Open-source Python framework for Cayley graphs and large algebraic structures; participation paused to focus on RL research. Details
- Gestalt Coupling Layer: Rethinking Intelligence as Resource Coordination — Concept note. Research direction on resource coordination in multi-agent systems. Details
Talks
No public talk recordings listed yet. Contact me for seminar or conference materials upon request.
Open Research Questions
- How to formalize operator uncertainty and approval latency in hierarchical HITL-RL without breaking policy improvement guarantees?
- Can evidential (DST) belief updates be integrated into constrained MDP decision layers for interpretable maintenance diagnostics?
- What resource-allocation mechanisms best support leadership migration in marginal hierarchical multi-agent control?
- How to evaluate simulation-based maintenance policies under distributional shift between synthetic environments and operational data?