active
Machine Repair Simulator
A scalable testbed for controlled machine-interference systems
This research-oriented testbed represents identical production machines that fail at speed-dependent rates, enter a FIFO repair queue, and return to operation after repair.
Research questions
The project examines when exact dynamic programming becomes computationally impractical, how state representations affect Markov sufficiency, and where approximate dynamic programming or reinforcement learning becomes attractive.
Methods
- SMDP
- Simulation
- Relative Value Iteration
- Reliability