![]() ![]() Giles, (Eds.), Advances in Neural Information Processing Systems 5. Hasselmo, (Eds.), Advances in Neural Information Processing Systems 8. Improving elevator performance using reinforcement learning. Proceedings of the Ninth Yale Workshop on Adaptive and Learning Systems.Ĭrites, R. Forming control policies from simulation models using reinforcement learning. PhD thesis, University of Massachusetts.Ĭrites, R.H. Large-Scale Dynamic Optimization Using Teams of Reinforcement Learning Agents. Homewood, IL: Aksen Associates.Ĭrites, R.H. ![]() Discrete Event Systems: Modeling and Performance Analysis. Cambridge, MA: MIT Press.Ĭassandras, C.G. ![]() Leen, (Eds.), Advances in Neural Information Processing Systems 7. Reinforcement learning methods for continuous-time Markov decision problems. Distributed adaptive optimal control of flexible structures. Belmont, MA: Athena Scientific Press.īradtke, S.J. Artificial Intelligence, 72, 81–138.īertsekas, D.P. ![]() Learning to act using real-time dynamic programming. Wokingham, England: Addison-Wesley.īarto, A.G., Bradtke, S.J., & Singh, S.P. From chemotaxis to cooperativity: Abstract exercises in neuronal learning strategies. ECE Department Technical Report, University of Massachusetts.īarto, A.G. Elevator dispatchers for down peak traffic. New York, NY: Basic Books.īao, G., Cassandras, C.G., Djaferis, T.E., Gandhi, A.D., & Looze, D.P. These results demonstrate the power of multi-agent RL on a very large scale stochastic dynamic optimization problem of practical utility.Īxelrod, R.M. In spite of these complications, we show results that in simulation surpass the best of the heuristic elevator control algorithms of which we are aware. The team receives a global reward signal which appears noisy to each agent due to the effects of the actions of the other agents, the random nature of the arrivals and the incomplete observation of the state. We use a team of RL agents, each of which is responsible for controlling one elevator car. It is a difficult domain posing a combination of challenges not seen in most multi-agent learning research to date. In this paper we demonstrate that such collective RL algorithms can be powerful heuristic methods for addressing large-scale control problems.Įlevator group control serves as our testbed. If each member of a team of agents employs one of these algorithms, a new collective learning algorithm emerges for the team as a whole. They can be trained on the basis of real or simulated experiences, focusing their computation on areas of state space that are actually visited during control, making them computationally tractable on very large problems. RL algorithms have appeared that approximate dynamic programming on an incremental basis. Recent algorithmic and theoretical advances in reinforcement learning (RL) have attracted widespread interest. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |