Successes Of RL

From RL-Community

Jump to: navigation, search

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

...

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
Personal tools