The Multiconference on Computational Engineering in Systems Applications

A Study of Multiagent Reinforcement Learning based on Quantum Theory

Meng Xiangping, Yuzhen Pi, Yuan Quande, Pan Ying

Published Oct 1, 2006• 3 citations

Abstract

In this paper, we present a novel multiagent reinforcement learning algorithm based on Q-learning and quantum theory. As in reinforcement learning algorithm, when the number of agents or/and agent's action is large enough, all of the action selection methods will be failed: the speed of learning is decreased sharply, we try to combine the quantum theory with Q-learning, hoping that the problem will be resolved with our proposed.

DOI: 10.1109/cesa.2006.4281965

Related papers

A Novel Multiagent Reinforcement Learning Algorithm Combination with Quantum Computation

Xiangping Meng, Yu Chen, Yuzhen Pi +1 more

A novel learning policy in multiagent reinforcement learning is presented, trying to find another tradeoff of exploration and exploitation efficiently, which is different from traditional greedy or softmax action selection method. The state and action of multiagent are represented with quantum superposition state, and probability amplitude is used to denote the probability of an action. Quantum search algorithm is adopted in multiagent action selection. The experiment results show that the new algorithm is effective and can help multiagent learn faster. This combination of quantum computing with multiagent reinforcement learning is an attempt, and the idea possibly brings more researches in multiagent reinforcement learning.

World Congress on Intelligent Control and AutomationJan 1, 20062 citations

A StudyofMultiagent Reinforcement Learning basedonQuantumTheory

Multiagent Q-Learning

Inthispaper, we present a novelMultiagent Reinforcement Learning Algorithm basedonQ-Learning and QuantumTheory.As in reinforcement learning algorithm, whenthenumberofagents or/and agent's action islarge enough, alloftheaction selection methods willbe failed: thespeedoflearning isdecreased sharply. wetryto combine thequantumtheory withQ-Learning, hoping that theproblem will beresolved withourproposed.

Jan 1, 20060 citations

Cooperative multi-agent reinforcement learning based on quantum computing

Xiangping Meng

Due to the interactions among the agents in the cooperative multi-agent systems,multi-agent learning problem complexity can rise rapidly with the number of agents or their behavioral sophistication.In order to converge to desirable equilibrium,agents generally require sufficient exploration of strategy space and coordinate their policies to achieve optimal equilibrium.A novel cooperative multi-agent learning method is proposed based on quantum theory and reinforcement learning.This method not only coordinates agents’ behaviors using quantum entanglement and helps agents make action selection under quantum superposition,but also adopts Grover’s searching algorithm which can speed up learning.This method also makes a good tradeoff between exploration and exploitation using probability characteristics of quantum theory.The results of simulated experiments show that quantum theory can be effectively used to cooperative multi-agent reinforcement learning.

Computer Engineering and Applications JournalJan 1, 20080 citations

Multi-agent reinforcement learning based on quantum chaotic computer

Xiangping Meng, Xinxin Wang, Jun Meng

A novel learning policy in multi-agent reinforcement learning is presented, trying to find another tradeoff of exploration and exploitation efficiently, It use the output of the classical quantum computer as an input for chaotic dynamics amplifier, The novel amplifier consider the chaotic effect, it can amplify the initial value in polynomial time. It considers the action selection problem and argues that the problem, in principle, can be solved in polynomial time if it combines the quantum computer with the chaotic dynamics amplifier based on the logistic map.

IEEE International Conference on Automation and Logistics (ICAL)Sep 1, 20080 citations

Quantum Chaotic Reinforcement Learning

Xiangping Meng, Jun Meng, Li-Juan Lui

International Conference on Natural Computation (ICNC)Jan 1, 20085 citations

Multi-Agent Q-learning based on quantum theory and ant colony algorithm

Wang Sheng-bin

Due to the interactions among the Agents in the cooperative multi-Agent systems,multi-Agent learning problem complexity can rise rapidly with the number of Agents or their behavioral sophistication.In order to converge to desirable equilibrium,Agents generally require sufficient exploration of strategy space and coordinate their policies to achieve optimal equilibrium.A novel cooperative multi-Agent learning method is proposed based on quantum theory,ant algorithm and Q-learning.First,this method not only coordinates Agents'behaviors using quantum entanglement and helps Agents make action selection under quantum superposition,but also adopts Grover's searching algorithm which can probe the action,speed up learn-ing.Second,according to ant algorithm,footmark thought is presented so that Agents can be indirectly enforced to communicate with others.At last,the theory analysis and result of experiment both demonstrate that the improved Q-learning is feasible and increases the learning efficiency.

Computer Engineering and Applications JournalJan 1, 20100 citations