Improving the speed of convergence of multi-agent Q-learning for cooperative task-planning by a robot-team

Sadhu, Arup Kumar; Konar, Amit

doi:10.1016/j.robot.2017.03.003

Cited by 30 publications

(14 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A comparative analysis of the proposed approach is done with other approaches mentioned in the literature. Ability and complexity (Sadhu and Konar 2017 ; Gadaleta 2017 ) both are analyzed with respect to the utilized time space and effectiveness. Time required to complete the execution process is referred as time complexity.…”

Section: Results Analysismentioning

confidence: 99%

Prevention of Covid-19 affected patient using multi robot cooperation and Q-learning approach: a solution

et al. 2021

View full text Add to dashboard Cite

Section: Results Analysismentioning

confidence: 99%

Prevention of Covid-19 affected patient using multi robot cooperation and Q-learning approach: a solution

et al. 2021

View full text Add to dashboard Cite

“…In most cases, Q-function with multistate is used to better learn the environment (Fernandez-Gauna et al, 2013;Sadhu and Konar, 2017;Chai and Hayashibe, 2020). Particularly, this paper (Chai and Hayashibe, 2020) has explored deep RL for motion generation in a simulated environment.…”

Section: Q-learning Structurementioning

confidence: 99%

Reinforcement Q-Learning Control With Reward Shaping Function for Swing Phase Control in a Semi-active Prosthetic Knee

et al. 2020

View full text Add to dashboard Cite

In this study, we investigated a control algorithm for a semi-active prosthetic knee based on reinforcement learning (RL). Model-free reinforcement Q-learning control with a reward shaping function was proposed as the voltage controller of a magnetorheological damper based on the prosthetic knee. The reward function was designed as a function of the performance index that accounts for the trajectory of the subject-specific knee angle. We compared our proposed reward function to a conventional single reward function under the same random initialization of a Q-matrix. We trained this control algorithm to adapt to several walking speed datasets under one control policy and subsequently compared its performance with that of other control algorithms. The results showed that our proposed reward function performed better than the conventional single reward function in terms of the normalized root mean squared error and also showed a faster convergence trend. Furthermore, our control strategy converged within our desired performance index and could adapt to several walking speeds. Our proposed control structure has also an overall better performance compared to user-adaptive control, while some of its walking speeds performed better than the neural network predictive control from existing studies.

show abstract

“…Busoniu et al [2] provides a comprehensive survey of multi-agent reinforcement learning (MARL) techniques for fully cooperative, fully competitive, and mixed tasks which focuses on autonomous multiple agents learning to solve complex tasks online using learning strategies based on dynamic programming and temporal-difference RL. Sadhu et al [3] proposed an algorithm for MAQL with two appealing features which were unavailable in convention Q learning. To begin with, an agent only needs to adapt one Q-table in joint state-action space during the learning process, as opposed to m joint Q-tables for, m agents in CQL.…”

Section: Introductionmentioning

confidence: 99%

“…This helps in the convergence of the process. Sadhu et al [3] has efficient convergence as it implements team-goal exploration and joint action selection for the specified state. Chaplot et al [6] focuses on Active Neural SLAM, a modular navigational model.…”

Section: Introductionmentioning

confidence: 99%

Development of Conceptual Framework for Reinforcement Learning Based Optimal Control

Gupta

Pal

Kumar

et al. 2021

Lecture Notes in Electrical Engineering

View full text Add to dashboard Cite

The framework of reinforcement learning-based optimal control depends on a mathematical formulation of intelligent decision making. In this article, we demonstrated the comprehensive design framework for offline reinforcement learning algorithms that utilizes sparse and discrete data space for efficient decisionmaking purposes. Learning is often difficult with the sparse reward function under the absence of optimization. Hence, an optimized map can be used in the reward function to improve efficacy. Some reward functions outperform sparse reward, such as "map completeness" and "information gain". "Map completeness" is proportional to the difference between the current time step and the previous time step, while "information gain" utilizes the entropy information for measuring the uncertainty in the map. On a whole. this article proposes a framework for the development of optimized reward function-based reinforcement learning based control strategy.

show abstract

Improving the speed of convergence of multi-agent Q-learning for cooperative task-planning by a robot-team

Cited by 30 publications

References 20 publications

Prevention of Covid-19 affected patient using multi robot cooperation and Q-learning approach: a solution

Prevention of Covid-19 affected patient using multi robot cooperation and Q-learning approach: a solution

Reinforcement Q-Learning Control With Reward Shaping Function for Swing Phase Control in a Semi-active Prosthetic Knee

Development of Conceptual Framework for Reinforcement Learning Based Optimal Control

Contact Info

Product

Resources

About