Academic literature on the topic 'Multi objective RL'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the lists of relevant articles, books, theses, conference reports, and other scholarly sources on the topic 'Multi objective RL.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Journal articles on the topic "Multi objective RL"
1
Ding, Li, and Lee Spector. "Multi-Objective Evolutionary Architecture Search for Parameterized Quantum Circuits." Entropy 25, no. 1 (January 3, 2023): 93. http://dx.doi.org/10.3390/e25010093.
Full textAbstract:
Recent work on hybrid quantum-classical machine learning systems has demonstrated success in utilizing parameterized quantum circuits (PQCs) to solve the challenging reinforcement learning (RL) tasks, with provable learning advantages over classical systems, e.g., deep neural networks. While existing work demonstrates and exploits the strength of PQC-based models, the design choices of PQC architectures and the interactions between different quantum circuits on learning tasks are generally underexplored. In this work, we introduce a Multi-objective Evolutionary Architecture Search framework for parameterized quantum circuits (MEAS-PQC), which uses a multi-objective genetic algorithm with quantum-specific configurations to perform efficient searching of optimal PQC architectures. Experimental results show that our method can find architectures that have superior learning performance on three benchmark RL tasks, and are also optimized for additional objectives including reductions in quantum noise and model size. Further analysis of patterns and probability distributions of quantum operations helps identify performance-critical design choices of hybrid quantum-classical learning systems.
2
Pianosi, F., A. Castelletti, and M. Restelli. "Tree-based fitted Q-iteration for multi-objective Markov decision processes in water resource management." Journal of Hydroinformatics 15, no. 2 (January 2, 2013): 258–70. http://dx.doi.org/10.2166/hydro.2013.169.
Full textAbstract:
Multi-objective Markov decision processes (MOMDPs) provide an effective modeling framework for decision-making problems involving water systems. The traditional approach is to define many single-objective problems (resulting from different combinations of the objectives), each solvable by standard optimization. This paper presents an approach based on reinforcement learning (RL) that can learn the operating policies for all combinations of objectives in a single training process. The key idea is to enlarge the approximation of the action-value function, which is performed by single-objective RL over the state-action space, to the space of the objectives' weights. The batch-mode nature of the algorithm allows for enriching the training dataset without further interaction with the controlled system. The approach is demonstrated on a numerical test case study and evaluated on a real-world application, the Hoa Binh reservoir, Vietnam. Experimental results on the test case show that the proposed approach (multi-objective fitted Q-iteration; MOFQI) becomes computationally preferable over the repeated application of its single-objective version (fitted Q-iteration; FQI) when evaluating more than five weight combinations. In the Hoa Binh case study, the operating policies computed with MOFQI and FQI have comparable efficiency, while MOFQI provides a continuous approximation of the Pareto frontier with no additional computing costs.
3
Wang, Yimeng, Mridul Agarwal, Tian Lan, and Vaneet Aggarwal. "Learning-Based Online QoE Optimization in Multi-Agent Video Streaming." Algorithms 15, no. 7 (June 28, 2022): 227. http://dx.doi.org/10.3390/a15070227.
Full textAbstract:
Video streaming has become a major usage scenario for the Internet. The growing popularity of new applications, such as 4K and 360-degree videos, mandates that network resources must be carefully apportioned among different users in order to achieve the optimal Quality of Experience (QoE) and fairness objectives. This results in a challenging online optimization problem, as networks grow increasingly complex and the relevant QoE objectives are often nonlinear functions. Recently, data-driven approaches, deep Reinforcement Learning (RL) in particular, have been successfully applied to network optimization problems by modeling them as Markov decision processes. However, existing RL algorithms involving multiple agents fail to address nonlinear objective functions on different agents’ rewards. To this end, we leverage MAPG-finite, a policy gradient algorithm designed for multi-agent learning problems with nonlinear objectives. It allows us to optimize bandwidth distributions among multiple agents and to maximize QoE and fairness objectives on video streaming rewards. Implementing the proposed algorithm, we compare the MAPG-finite strategy with a number of baselines, including static, adaptive, and single-agent learning policies. The numerical results show that MAPG-finite significantly outperforms the baseline strategies with respect to different objective functions and in various settings, including both constant and adaptive bitrate videos. Specifically, our MAPG-finite algorithm maximizes QoE by 15.27% and maximizes fairness by 22.47% compared to the standard SARSA algorithm for a 2000 KB/s link.
4
Saksirinukul, Thanis, Permyot Kosolbhand, and Natthaporn Tanpowpong. "Increasing the remnant liver volume using portal vein embolization." Asian Biomedicine 4, no. 5 (October 1, 2010): 817–20. http://dx.doi.org/10.2478/abm-2010-0107.
Full textAbstract:
Abstract Background: Portal vein embolization (PVE) is a common procedure to induce hypertrophy of the remnant liver (RL) before major hepatectomy. Objective: Evaluate increased RL volume after PVE based on CT volumetric measurement. Methods: Multi-detector computed tomography (MDCT) was used to measure hepatic volumetric measurement, including total liver volume and RL volumes of pre- and post-PVE. Complications were recorded from PVE and from three-month after post-extended hepatectomy liver dysfunction. Result and conclusion: There was a 10% increase in RL volume. Mean days between CT and PVE were 20 days. No major complications from PVE were observed.
5
ZHANG, ZHICONG, WEIPING WANG, SHOUYAN ZHONG, and KAISHUN HU. "FLOW SHOP SCHEDULING WITH REINFORCEMENT LEARNING." Asia-Pacific Journal of Operational Research 30, no. 05 (October 2013): 1350014. http://dx.doi.org/10.1142/s0217595913500140.
Full textAbstract:
Reinforcement learning (RL) is a state or action value based machine learning method which solves large-scale multi-stage decision problems such as Markov Decision Process (MDP) and Semi-Markov Decision Process (SMDP) problems. We minimize the makespan of flow shop scheduling problems with an RL algorithm. We convert flow shop scheduling problems into SMDPs by constructing elaborate state features, actions and the reward function. Minimizing the accumulated reward is equivalent to minimizing the schedule objective function. We apply on-line TD(λ) algorithm with linear gradient-descent function approximation to solve the SMDPs. To examine the performance of the proposed RL algorithm, computational experiments are conducted on benchmarking problems in comparison with other scheduling methods. The experimental results support the efficiency of the proposed algorithm and illustrate that the RL approach is a promising computational approach for flow shop scheduling problems worthy of further investigation.
6
García, Javier, Roberto Iglesias, Miguel A. Rodríguez, and Carlos V. Regueiro. "Directed Exploration in Black-Box Optimization for Multi-Objective Reinforcement Learning." International Journal of Information Technology & Decision Making 18, no. 03 (May 2019): 1045–82. http://dx.doi.org/10.1142/s0219622019500093.
Full textAbstract:
Usually, real-world problems involve the optimization of multiple, possibly conflicting, objectives. These problems may be addressed by Multi-objective Reinforcement learning (MORL) techniques. MORL is a generalization of standard Reinforcement Learning (RL) where the single reward signal is extended to multiple signals, in particular, one for each objective. MORL is the process of learning policies that optimize multiple objectives simultaneously. In these problems, the use of directional/gradient information can be useful to guide the exploration to better and better behaviors. However, traditional policy-gradient approaches have two main drawbacks: they require the use of a batch of episodes to properly estimate the gradient information (reducing in this way the learning speed), and they use stochastic policies which could have a disastrous impact on the safety of the learning system. In this paper, we present a novel population-based MORL algorithm for problems in which the underlying objectives are reasonably smooth. It presents two main characteristics: fast computation of the gradient information for each objective through the use of neighboring solutions, and the use of this information to carry out a geometric partition of the search space and thus direct the exploration to promising areas. Finally, the algorithm is evaluated and compared to policy gradient MORL algorithms on different multi-objective problems: the water reservoir and the biped walking problem (the latter both on simulation and on a real robot).
7
Sharma, S. K., S. S. Mahapatra, and M. B. Parappagoudar. "Benchmarking of product recovery alternatives in reverse logistics." Benchmarking: An International Journal 23, no. 2 (March 7, 2016): 406–24. http://dx.doi.org/10.1108/bij-01-2014-0002.
Full textAbstract:
Purpose – Selection of best product recovery alternative in reverse logistics (RL) has gained great attention in supply chain community. The purpose of this paper is to provide a robust group decision-making tool to select the best product recovery alternative. Design/methodology/approach – In this paper, fuzzy values, assigned to various criteria and alternatives by a number of decision makers, are converted into crisp values and then aggregated scores are evaluated. After obtaining experts’ scores, objective and subjective weights of the criteria have been calculated using variance method and analytic hierarchy process, respectively. Then integrated weights of criteria are evaluated using different proportions of the two weights. The superiority and inferiority ranking (SIR) method is then employed to achieve the final ranking of alternatives. An example is presented to demonstrate the methodology. Findings – The proposed methodology provides decision makers a systematic, flexible and realistic approach to effectively rank the product recovery alternatives in RL. The alternatives can easily be benchmarked and best recovery strategy can be obtained. The sensitivity analysis carried out by changing different proportion of objective and subjective weights reveals that best ranking alternative never changes and proves the robustness of the methodology. The present benchmarking framework can also be used by decision makers to simplify any problem which encounters multi-attribute decision making and multiple decision makers. Research limitations/implications – The proposed methodology should be tested in different situations having varied operational and environmental conditions dealing with different products. A real case study from an industrial set up can help to assess the behavior of the proposed methodology. The presented methodology however can deal with such multi-disciplinary and multi-criteria issues in a simple and structured manner and ease the managers to select the best alternative. Originality/value – A novel approach for decision making taking into account both objective and subjective weights for criteria has been proposed to rank the best recovery alternatives in RL. The proposed methodology uses SIR method to prioritize the alternatives. As RL alternative selection is an important issue and involves both technical and managerial criteria as well as multiple decision makers, the proposed robust methodology can provide guidelines for the practicing managers.
8
Ren, Jianfeng, Chunming Ye, and Yan Li. "A Two-Stage Optimization Algorithm for Multi-objective Job-Shop Scheduling Problem Considering Job Transport." Journal Européen des Systèmes Automatisés 53, no. 6 (December 23, 2020): 915–24. http://dx.doi.org/10.18280/jesa.530617.
Full textAbstract:
This paper solves the job-shop scheduling problem (JSP) considering job transport, with the aim to minimize the maximum makespan, tardiness, and energy consumption. In the first stage, the improved fast elitist nondominated sorting genetic algorithm II (INSGA-II) was combined with N5 neighborhood structure and the local search strategy of nondominant relationship to generate new neighborhood solutions by exchanging the operations on the key paths. In the second stage, the ant colony algorithm based on reinforcement learning (RL-ACA) was designed to optimize the job transport task, abstract the task into polar coordinates, and further optimizes the task. The proposed two-stage algorithm was tested on small, medium, and large-scale examples. The results show that our algorithm is superior to other algorithms in solving similar problems.
9
Ramezani Dooraki, Amir, and Deok-Jin Lee. "A Multi-Objective Reinforcement Learning Based Controller for Autonomous Navigation in Challenging Environments." Machines 10, no. 7 (June 22, 2022): 500. http://dx.doi.org/10.3390/machines10070500.
Full textAbstract:
In this paper, we introduce a self-trained controller for autonomous navigation in static and dynamic (with moving walls and nets) challenging environments (including trees, nets, windows, and pipe) using deep reinforcement learning, simultaneously trained using multiple rewards. We train our RL algorithm in a multi-objective way. Our algorithm learns to generate continuous action for controlling the UAV. Our algorithm aims to generate waypoints for the UAV in such a way as to reach a goal area (shown by an RGB image) while avoiding static and dynamic obstacles. In this text, we use the RGB-D image as the input for the algorithm, and it learns to control the UAV in 3-DoF (x, y, and z). We train our robot in environments simulated by Gazebo sim. For communication between our algorithm and the simulated environments, we use the robot operating system. Finally, we visualize the trajectories generated by our trained algorithms using several methods and illustrate our results that clearly show our algorithm’s capability in learning to maximize the defined multi-objective reward.
10
Aaltonen, Harri, Seppo Sierla, Ville Kyrki, Mahdi Pourakbari-Kasmaei, and Valeriy Vyatkin. "Bidding a Battery on Electricity Markets and Minimizing Battery Aging Costs: A Reinforcement Learning Approach." Energies 15, no. 14 (July 6, 2022): 4960. http://dx.doi.org/10.3390/en15144960.
Full textAbstract:
Battery storage is emerging as a key component of intelligent green electricitiy systems. The battery is monetized through market participation, which usually involves bidding. Bidding is a multi-objective optimization problem, involving targets such as maximizing market compensation and minimizing penalties for failing to provide the service and costs for battery aging. In this article, battery participation is investigated on primary frequency reserve markets. Reinforcement learning is applied for the optimization. In previous research, only simplified formulations of battery aging have been used in the reinforcement learning formulation, so it is unclear how the optimizer would perform with a real battery. In this article, a physics-based battery aging model is used to assess the aging. The contribution of this article is a methodology involving a realistic battery simulation to assess the performance of the trained RL agent with respect to battery aging in order to inform the selection of the weighting of the aging term in the RL reward formula. The RL agent performs day-ahead bidding on the Finnish Frequency Containment Reserves for Normal Operation market, with the objective of maximizing market compensation, minimizing market penalties and minimizing aging costs.
Book chapters on the topic "Multi objective RL"
1
Xi, Wei, and Xian Guo. "Multi-objective RL with Preference Exploration." In Intelligent Robotics and Applications, 669–80. Cham: Springer International Publishing, 2022. http://dx.doi.org/10.1007/978-3-031-13844-7_62.
Full text
2
Hasan, Md Mahmudul, Md Shahinur Rahman, and Adrian Bell. "Deep Reinforcement Learning for Optimization." In Research Anthology on Artificial Intelligence Applications in Security, 1598–614. IGI Global, 2021. http://dx.doi.org/10.4018/978-1-7998-7705-9.ch070.
Full textAbstract:
Deep reinforcement learning (DRL) has transformed the field of artificial intelligence (AI) especially after the success of Google DeepMind. This branch of machine learning epitomizes a step toward building autonomous systems by understanding of the visual world. Deep reinforcement learning (RL) is currently applied to different sorts of problems that were previously obstinate. In this chapter, at first, the authors started with an introduction of the general field of RL and Markov decision process (MDP). Then, they clarified the common DRL framework and the necessary components RL settings. Moreover, they analyzed the stochastic gradient descent (SGD)-based optimizers such as ADAM and a non-specific multi-policy selection mechanism in a multi-objective Markov decision process. In this chapter, the authors also included the comparison for different Deep Q networks. In conclusion, they describe several challenges and trends in research within the deep reinforcement learning field.
3
Hasan, Md Mahmudul, Md Shahinur Rahman, and Adrian Bell. "Deep Reinforcement Learning for Optimization." In Research Anthology on Artificial Intelligence Applications in Security, 1598–614. IGI Global, 2021. http://dx.doi.org/10.4018/978-1-7998-7705-9.ch070.
Full textAbstract:
Deep reinforcement learning (DRL) has transformed the field of artificial intelligence (AI) especially after the success of Google DeepMind. This branch of machine learning epitomizes a step toward building autonomous systems by understanding of the visual world. Deep reinforcement learning (RL) is currently applied to different sorts of problems that were previously obstinate. In this chapter, at first, the authors started with an introduction of the general field of RL and Markov decision process (MDP). Then, they clarified the common DRL framework and the necessary components RL settings. Moreover, they analyzed the stochastic gradient descent (SGD)-based optimizers such as ADAM and a non-specific multi-policy selection mechanism in a multi-objective Markov decision process. In this chapter, the authors also included the comparison for different Deep Q networks. In conclusion, they describe several challenges and trends in research within the deep reinforcement learning field.
4
Hasan, Md Mahmudul, Md Shahinur Rahman, and Adrian Bell. "Deep Reinforcement Learning for Optimization." In Handbook of Research on Deep Learning Innovations and Trends, 180–96. IGI Global, 2019. http://dx.doi.org/10.4018/978-1-5225-7862-8.ch011.
Full textAbstract:
Deep reinforcement learning (DRL) has transformed the field of artificial intelligence (AI) especially after the success of Google DeepMind. This branch of machine learning epitomizes a step toward building autonomous systems by understanding of the visual world. Deep reinforcement learning (RL) is currently applied to different sorts of problems that were previously obstinate. In this chapter, at first, the authors started with an introduction of the general field of RL and Markov decision process (MDP). Then, they clarified the common DRL framework and the necessary components RL settings. Moreover, they analyzed the stochastic gradient descent (SGD)-based optimizers such as ADAM and a non-specific multi-policy selection mechanism in a multi-objective Markov decision process. In this chapter, the authors also included the comparison for different Deep Q networks. In conclusion, they describe several challenges and trends in research within the deep reinforcement learning field.
Conference papers on the topic "Multi objective RL"
1
Handa, Hisashi. "Solving Multi-objective Reinforcement Learning Problems by EDA-RL - Acquisition of Various Strategies." In 2009 Ninth International Conference on Intelligent Systems Design and Applications. IEEE, 2009. http://dx.doi.org/10.1109/isda.2009.92.
Full text
2
Dworschak, Fabian, Christopher Sauer, Benjamin Schleich, and Sandro Wartzack. "Reinforcement Learning As an Alternative for Parameter Prediction In Design for Sheet Bulk Metal Forming." In ASME 2022 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference. American Society of Mechanical Engineers, 2022. http://dx.doi.org/10.1115/detc2022-89073.
Full textAbstract:
Abstract This contribution presents an approach and a case study to compare Reinforcement Learning (RL) and Genetic Algorithms (GA) for parameter prediction in Sheet Bulk Metal Forming (SBMF). Machine Learning (ML) and Multi-Objective Optimization (MOO) to provide different points of view for the prediction of manufacturing parameters. While supervised learning depends on sufficient training data, GA lack the ability to explain how sufficient parameters were achieved. RL could help to overcome both issues, as it is independent from training data and can be used to learn a policy leading towards suitable parameter combinations, which can be tracked through the solution space. To probe RL in terms of feasibility for parameter prediction and necessary training effort SBMF serves as an appropriate use case because solution and objective function space are multidimensional, and their relations are challenging for MOO. The results of a Reinforcement Learner and a GA are compared and discussed to answer the question under which circumstances RL can provide an alternative for parameter prediction.
3
Tian, Zheng, Ying Wen, Zhichen Gong, Faiz Punakkath, Shihao Zou, and Jun Wang. "A Regularized Opponent Model with Maximum Entropy Objective." In Twenty-Eighth International Joint Conference on Artificial Intelligence {IJCAI-19}. California: International Joint Conferences on Artificial Intelligence Organization, 2019. http://dx.doi.org/10.24963/ijcai.2019/85.
Full textAbstract:
In a single-agent setting, reinforcement learning (RL) tasks can be cast into an inference problem by introducing a binary random variable o, which stands for the "optimality". In this paper, we redefine the binary random variable o in multi-agent setting and formalize multi-agent reinforcement learning (MARL) as probabilistic inference. We derive a variational lower bound of the likelihood of achieving the optimality and name it as Regularized Opponent Model with Maximum Entropy Objective (ROMMEO). From ROMMEO, we present a novel perspective on opponent modeling and show how it can improve the performance of training agents theoretically and empirically in cooperative games. To optimize ROMMEO, we first introduce a tabular Q-iteration method ROMMEO-Q with proof of convergence. We extend the exact algorithm to complex environments by proposing an approximate version, ROMMEO-AC. We evaluate these two algorithms on the challenging iterated matrix game and differential game respectively and show that they can outperform strong MARL baselines.
4
Mei, Kai, and Yilin Fang. "Multi-Robotic Disassembly Line Balancing Using Deep Reinforcement Learning." In ASME 2021 16th International Manufacturing Science and Engineering Conference. American Society of Mechanical Engineers, 2021. http://dx.doi.org/10.1115/msec2021-63522.
Full textAbstract:
Abstract With the continuous development of society, the speed of product upgrades is getting faster and faster, and the recycling of EOL products has great benefits to environmental protection and resource utilization. In recent years, with the rapid development of artificial intelligence (AI), more and more scholars have begun to apply reinforcement learning (RL) and deep learning (DL) to solve practical problems. This paper focuses on the application of deep reinforcement learning (DRL) in the multi-robotic disassembly line balance problem (MRDLBP). In the MRDLBP problem, for each workstation, the cycle time (CT) is determined, and the robot resources that can be accommodated are optional. Multi-objective includes minimizing workstation idle time, priority disassembly of high-demand components and minimizing energy consumption. The input model is single, and robotic resources are variable. Firstly, we formulated the mathematical model of the problem and proposed a framework for MRDLBP using DRL. In addition, we modeled the DRL system with three DRL algorithms, including deep Q network (DQN), double DQN (D_DQN) and prioritized experience replay DQN (PRDQN), to solve the problem. Finally, we build different test cases by adjusting the type of input model and the number of robot resources to test the performance of the three algorithms under different complexity conditions.
5
Bhowmik, Subrata. "Machine Learning-Based Optimization for Subsea Pipeline Route Design." In Offshore Technology Conference. OTC, 2021. http://dx.doi.org/10.4043/31031-ms.
Full textAbstract:
Abstract Optimal route selection for the subsea pipeline is a critical task for the pipeline design process, and the route selected can significantly affect the overall project cost. Therefore, it is necessary to design the routes to be economical and safe. On-bottom stability (OBS) and fixed obstacles like existing crossings and free spans are the main factors that affect the route selection. This article proposes a novel hybrid optimization method based on a typical Machine Learning algorithm for designing an optimal pipeline route. The proposed optimal route design is compared with one of the popular multi-objective optimization method named Genetic Algorithm (GA). The proposed pipeline route selection method uses a Reinforcement Learning (RL) algorithm, a particular type of machine learning method to train a pipeline system that would optimize the route selection of subsea pipelines. The route optimization tool evaluates each possible route by incorporating Onbottom stability criteria based on DNVGL-ST-109 standard and other constraints such as the minimum pipeline route length, static obstacles, pipeline crossings, and free-span section length. The cost function in the optimization method simultaneously handles the minimization of length and cost of mitigating procedures. Genetic Algorithm, a well established optimization method, has been used as a reference to compare the optimal route with the result from the proposed Reinforcement Learning based optimization method. Three different case studies are performed for finding the optimal route selection using the Reinforcement Learning (RL) approach considering the OBS criteria into its cost function and compared with the Genetic Algorithm (GA). The RL method saves upto 20% pipeline length for a complex problem with 15 crossings and 31 free spans. The RL optimization method provides the optimal routes, considering different aspects of the design and the costs associated with the various factors to stabilize a pipeline (mattress, trenching, burying, concrete coating, or even employing a more massive pipe with additional steel wall thickness). OBS criteria significantly influence the best route, indicating that the tool can reduce the pipeline's design time and minimize installation and operational costs of the pipeline. Conventionally the pipeline route optimization is performed by a manual process where the minimum roule length and static obstacles are considered to find an optimum route. The engineering is then performed to fulfill the criteria of this route, and this approach may not lead to an optimized engineering cost. The proposed Reinforced Learning method for route optimization is a mixed type, faster, and cost-efficient approach. It significantly minimizes the pipeline's installation and operational costs up to 20% of the conventional route selection process.