Safe reinforcement learning with online filtering for fatigue-predictive human-robot task planning and allocation in production

2026-04-14 · Source: Artificial Intelligence · Field: Manufacturing & Industrial — Automation & Robotics, Artificial Intelligence & Machine Learning, Manufacturing Operations & Management · Depth: Expert, quick

Summary

A new safe reinforcement learning approach, PF-CD3Q, has been developed to address dynamic human-robot task planning and allocation (HRTPA) in manufacturing, a key component of Industry 5.0. This method aims to maximize production efficiency while ensuring worker physical fatigue remains within safe limits. Unlike traditional HRTPA models that use static fatigue parameters, PF-CD3Q accounts for daily variations in human fatigue sensitivity by estimating fatigue-related parameters online. It integrates a particle filter (PF) for real-time tracking of human fatigue and updating fatigue model parameters. These estimators are then incorporated into a constrained dueling double deep Q-learning (CD3Q) framework, which predicts task-level fatigue and excludes tasks exceeding fatigue limits, effectively formulating the problem as a constrained Markov decision process (CMDP).

Key takeaway

For research scientists developing human-robot collaboration systems, PF-CD3Q offers a robust method to incorporate dynamic human fatigue into task planning. You should consider implementing online fatigue parameter estimation to enhance the adaptability and safety of your HRTPA models, moving beyond static assumptions to better reflect real-world worker conditions and improve overall system reliability.

Key insights

PF-CD3Q uses safe reinforcement learning and online fatigue estimation for dynamic human-robot task allocation.

Principles

• Ergonomics enhances worker well-being.
• Fatigue parameters vary daily.
• Online estimation improves model accuracy.

Method

PF-CD3Q integrates particle filter (PF) estimators for real-time fatigue tracking and parameter updates into a constrained dueling double deep Q-learning (CD3Q) framework, predicting fatigue and constraining action space.

In practice

• Apply online fatigue estimation.
• Use constrained RL for safety.
• Integrate PF with deep Q-learning.

Topics

• Safe Reinforcement Learning
• Human-Robot Collaboration
• Task Planning and Allocation
• Fatigue Prediction
• Particle Filter

Best for: Research Scientist, Robotics Engineer, Machine Learning Engineer, AI Scientist

Related on AIssential

• A hierarchical spatial-aware algorithm with efficient reinforcement learning for human-robot task planning and allocation in production — A hierarchical reinforcement learning algorithm optimizes human-robot task planning by integrating spatial awareness and efficient deep Q-learning.
• RL-STPA: Adapting System-Theoretic Hazard Analysis for Safety-Critical Reinforcement Learning — RL-STPA systematically identifies and mitigates hazards in safety-critical RL systems through adapted STPA principles.
• Provable Offline Reinforcement Learning for Structured Cyclic MDPs — CycleFQI offers a modular, theoretically robust offline RL solution for cyclic MDPs with heterogeneous stages.
• Reinforcement Learning with Human Feedback (RLHF) in 4 minutes — RLHF uses human feedback and reinforcement learning to align LLM outputs with human preferences.
• Safe Deep Reinforcement Learning for Building Heating Control and Demand-side Flexibility — A real-time adaptive safety filter enhances DRL for building heating, ensuring demand-side flexibility compliance and efficiency.
• Reinforcement Learning Foundation Models Should Already Be A Thing — Reinforcement Learning (RL) can leverage synthetic data and fixed-size MDP statistics to create foundation models, akin to language and vision.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.