Successes Of RL
From RL-Community
This page was proposed on the RL-List and members of the community seemed eager to jump in and create a resource listing some of the theoretical, simulated, and empirical successes of reinforcement learning! Feel free to start filling it in!
ATTENTION: This page is under construction and is not finished yet. That doesn't mean that there are no successes in RL!
Contents |
Classics
- TD-Gammon
- Elevator control
- Dynamic Catalog Mailing
- Dialogue Management
- Distributed Network Routing
- Call Admission Control
- Job-shop scheduling
...
Theoretical Success
Success in Simulation
Empirical Success
Links
RL Applications page on Csaba Szepesvári's website, listing papers describing recent engineering/OR applications (+ a bunch of random others); the corresponding bib file Some of he application areas that come up in this list include:
- Networking (dynamic channel allocation, call admission control, QoS routing, rate control, call routing in call centers, underwater routing, power control, adaptive waveform selection in cognitive radio, ..)
- Taxi-out time prediction for aircrafts
- Recovering the components of a signal
- OR problems: Supply chain management, inventory control, asset management, airline yield management, preventive maintenance schedule optimization
- Sensor networks (sensor management)
- Autonomic management (web service optimization, load balancing, grid scheduling, power control, adaptive algorithms)
- Traffic control (urban, freeway, train traffic disturbance rejection)
- Control of automotive engines (fuel efficiency, hybrid vehicles, ..)
- Robotics (visual navigation, control of snake-like robots, autonomous underwater cable tracking, micromanipulation, path tracking, vibration isolation, learning movement primitives, stair climbing gait for humanoids, RoboCup keepaway soccer, ..)
- Aircraft autolander, morphing control of aircraft, helicopter control, altitude control of aircraft, Temperature profile control of a high-speed aerospace vehicle
- Natural language dialogue management
- Process control (bioreactors, chemical reactors)
- Object recognition - control of search
- writing jobs online the original essays
- Pricing of financial derivatives
- Power systems (damping of oscillations, economic power dispatching, )
- Training of brain-machine interfaces
- Natural language processing (more generally, discriminative fitting of generative models for structured prediction problems)
- Disaster policy optimization
- Optimization of off-chip DRAM bandwidth utilization
- Learning time allocation in games
- Gas turbine control
- Andrew Ng's page has a number of cool applications
- Warren Powell and his CASTLE lab is looking at various applications in operations research
- Peter Stone's page (RoboCup,..)
- UCT, a planning algorithm coming from the RL community revolutionarizes how computers play Go
- Real cart-pole balancing with a cool video; the RL algorithm learns in 2 and a half minute
- Satinder Singh's similar page on Successes of RL
