Automated Patient-Robot Task Assignment in a Simulated Stochastic Rehabilitation Gym
Summary
This study introduces an automated system for assigning patients to rehabilitation robots to maximize skill development. The system uses a neural network in a stochastic environment, outperforming fixed or limited robot-switching schedules.
Area of Science:
- Robotics
- Neurorehabilitation
- Machine Learning
Background:
- Robotic systems enhance rehabilitation after neurological injury.
- Rehabilitation robot gyms allow simultaneous patient training on diverse exercises.
- Current multipatient supervision lacks optimal automated patient-robot assignment.
Purpose of the Study:
- To develop an automated assignment system for maximizing patient skill development in a stochastic rehabilitation environment.
- To improve upon previous deterministic assignment models by incorporating randomness and predictive skill estimation.
- To enhance multipatient supervision through intelligent robot allocation.
Main Methods:
- Developed a stochastic environment simulating patient-robot interactions with random skill development components.
- Implemented a neural network to estimate patient skill development based on historical training success rates.
- Utilized a timestep-by-timestep assignment strategy to maximize group skill development.
Main Results:
- The proposed automated assignment system generated schedules that outperformed baseline strategies.
- Baseline strategies included fixed patient-robot assignments and single mid-session robot switches.
- Simulation trials demonstrated the effectiveness of the adaptive assignment approach.
Conclusions:
- Automated, adaptive patient-robot assignment systems can significantly enhance rehabilitation outcomes.
- Neural network-based skill prediction is crucial for optimizing assignments in stochastic environments.
- The developed system offers a promising approach for improving multipatient robotic rehabilitation supervision.
