Search results for: multi-agent reinforcement learning (MARL)

Authors: Sirapop Nuannimnoi, Hojjat Baghban, Ching-Yao Huang

Abstract:

Over the past decade, with the fast development of blockchain technology since the birth of Bitcoin, there has been a massive increase in the usage of Cryptocurrencies. Cryptocurrencies are not seen as an investment opportunity due to the market’s erratic behavior and high price volatility. With the recent success of deep reinforcement learning (DRL), portfolio management can be modeled and automated. In this paper, we propose a novel DRL-based multi-agent system to automatically make proper trading decisions on multiple cryptocurrencies and gain profits in the highly volatile cryptocurrency market. We also extend this multi-agent system with horizontal federated transfer learning for better adapting to the inclusion of new cryptocurrencies in our portfolio; therefore, we can, through the concept of diversification, maximize our profits and minimize the trading risks. Experimental results through multiple simulation scenarios reveal that this proposed algorithmic trading system can offer three promising key advantages over other systems, including maximized profits, minimized risks, and adaptability.

Keywords: cryptocurrency portfolio management, algorithmic trading, federated learning, multi-agent reinforcement learning

Procedia PDF Downloads 112

Authors: Yihao Kuang, Bowen Ding

Abstract:

With the increasing scale and complexity of knowledge graph, modern knowledge graph contains more and more types of entity, relationship, and attribute information. Therefore, in recent years, it has been a trend for knowledge graph inference to use reinforcement learning to deal with large-scale, incomplete, and noisy knowledge graphs and improve the inference effect and interpretability. The Proximal Policy Optimization (PPO) algorithm utilizes a near-end strategy optimization approach. This allows for more extensive updates of policy parameters while constraining the update extent to maintain training stability. This characteristic enables PPOs to converge to improved strategies more rapidly, often demonstrating enhanced performance early in the training process. Furthermore, PPO has the advantage of offline learning, effectively utilizing historical experience data for training and enhancing sample utilization. This means that even with limited resources, PPOs can efficiently train for reinforcement learning tasks. Based on these characteristics, this paper aims to obtain a better and more efficient inference effect by introducing PPO into knowledge inference technology.

Keywords: reinforcement learning, PPO, knowledge inference

Procedia PDF Downloads 229

Authors: Brwa Abdulrahman Abubaker

Abstract:

Recently, there has been a lot of interest in the difficult task of applying reinforcement learning to autonomous mobile robots. Conventional reinforcement learning (TRL) techniques have many drawbacks, such as lengthy computation times, intricate control frameworks, a great deal of trial and error searching, and sluggish convergence. In this paper, a modified Spiking Neural Network (SNN) is used to offer a distinct method for autonomous mobile robot learning and control in unexpected surroundings. As a learning algorithm, the suggested model combines dopamine modulation with spike-timing-dependent plasticity (STDP). In order to create more computationally efficient, biologically inspired control systems that are adaptable to changing settings, this work uses the effective and physiologically credible Izhikevich neuron model. This study is primarily focused on creating an algorithm for target tracking in the presence of obstacles. Results show that the SNN trained with three obstacles yielded an impressive 96% success rate for our proposal, with collisions happening in about 4% of the 214 simulated seconds.

Keywords: spiking neural network, spike-timing-dependent plasticity, dopamine modulation, reinforcement learning

Procedia PDF Downloads 5

Authors: Ali Nadi, Ali Edrissi

Abstract:

Relief demand and transportation links availability is the essential information that is needed for every natural disaster operation. This information is not in hand once a disaster strikes. Relief demand and network condition has been evaluated based on prediction method in related works. Nevertheless, prediction seems to be over or under estimated due to uncertainties and may lead to a failure operation. Therefore, in this paper a stochastic programming model is proposed to evaluate real-time relief demand and network condition at the onset of a natural disaster. To address the time sensitivity of the emergency response, the proposed model uses reinforcement learning for optimization of the total relief assessment time. The proposed model is tested on a real size network problem. The simulation results indicate that the proposed model performs well in the case of collecting real-time information.

Keywords: disaster management, real-time demand, reinforcement learning, relief demand

Procedia PDF Downloads 307

Authors: Shweta Singh, Sudaman Katti

Abstract:

The most well-known sequence models make use of complex recurrent neural networks in an encoder-decoder configuration. The model used in this research makes use of a transformer, which is based purely on a self-attention mechanism, without relying on recurrence at all. More specifically, encoders and decoders which make use of self-attention and operate based on a memory, are used. In this research work, results for various 3D visual and non-visual reinforcement learning tasks designed in Unity software were obtained. Convolutional neural networks, more specifically, nature CNN architecture, are used for input processing in visual tasks, and comparison with standard long short-term memory (LSTM) architecture is performed for both visual tasks based on CNNs and non-visual tasks based on coordinate inputs. This research work combines the transformer architecture with the proximal policy optimization technique used popularly in reinforcement learning for stability and better policy updates while training, especially for continuous action spaces, which are used in this research work. Certain tasks in this paper are long horizon tasks that carry on for a longer duration and require extensive use of memory-based functionalities like storage of experiences and choosing appropriate actions based on recall. The transformer, which makes use of memory and self-attention mechanism in an encoder-decoder configuration proved to have better performance when compared to LSTM in terms of exploration and rewards achieved. Such memory based architectures can be used extensively in the field of cognitive robotics and reinforcement learning.

Keywords: convolutional neural networks, reinforcement learning, self-attention, transformers, unity

Procedia PDF Downloads 127

Authors: G. Valentis, Q. Berthelot

Abstract:

Nowadays the trend to develop ever more intelligent and autonomous systems often takes its inspiration in the living beings on Earth. Some simple isolated systems are able, once brought together, to form a strong and reliable system. When trying to adapt the idea to man-made systems it is not possible to include in their program everything the system may encounter during its life cycle. It is, thus, necessary to make the system able to take decisions based on other criteria such as its past experience, i.e. to make the system learn on its own. However, at some point the acquired knowledge depends also on environment. So the question is: if system environment is modified, how could the system respond to it quickly and appropriately enough? Here, starting from reinforcement learning to rate its decisions, and using adaptive learning algorithms for gain and loss reward, the system is made able to respond to changing environment and to adapt its knowledge as time passes. Application is made to a robot finding an exit in a labyrinth.

Keywords: reinforcement learning, neural network, autonomous systems, adaptive learning, changing environment

Procedia PDF Downloads 418

Authors: Niklas Panten, Eberhard Abele

Abstract:

This paper presents a novel approach for real-time and near-optimal control of industrial smart grids by deep reinforcement learning (DRL). To achieve highly energy-efficient factory systems, the energetic linkage of machines, technical building equipment and the building itself is desirable. However, the increased complexity of the interacting sub-systems, multiple time-variant target values and stochastic influences by the production environment, weather and energy markets make it difficult to efficiently control the energy production, storage and consumption in the hybrid industrial smart grids. The studied deep reinforcement learning approach allows to explore the solution space for proper control policies which minimize a cost function. The deep neural network of the DRL agent is based on a multilayer perceptron (MLP), Long Short-Term Memory (LSTM) and convolutional layers. The agent is trained within multiple Modelica-based factory simulation environments by the Advantage Actor Critic algorithm (A2C). The DRL controller is evaluated by means of the simulation and then compared to a conventional, rule-based approach. Finally, the results indicate that the DRL approach is able to improve the control performance and significantly reduce energy respectively operating costs of industrial smart grids.

Keywords: industrial smart grids, energy efficiency, deep reinforcement learning, optimal control

Procedia PDF Downloads 183

Authors: Nitin Singh, Meng Ling, Talha Ahmed, Tianxia Zhao, Reinier van de Pol

Abstract:

We propose the use of reinforcement learning (RL) as a viable alternative for optimizing supply chain management, particularly in scenarios with stochasticity in product demands. RL’s adaptability to changing conditions and its demonstrated success in diverse fields of sequential decision-making makes it a promising candidate for addressing supply chain problems. We investigate the impact of demand fluctuations in a multi-product supply chain system and develop RL agents with learned generalizable policies. We provide experimentation details for training RL agents and statistical analysis of the results. We study the generalization ability of RL agents for different demand uncertainty scenarios and observe superior performance compared to the agents trained with fixed demand curves. The proposed methodology has the potential to lead to cost reduction and increased profit for companies dealing with frequent inventory movement between supply and demand nodes.

Keywords: inventory management, reinforcement learning, supply chain optimization, uncertainty

Procedia PDF Downloads 102

Authors: Akshay Paranjape, Nils Plettenberg, Robert Schmitt

Abstract:

Producing faulty products can be costly for manufacturing companies and wastes resources. To reduce scrap rates in manufacturing, process parameters can be optimized using machine learning. Thus far, research mainly focused on optimizing specific processes using traditional algorithms. To develop a framework that enables real-time optimization based on a predictive model for an arbitrary production process, this study explores the application of reinforcement learning (RL) in this field. Based on a thorough review of literature about RL and process parameter optimization, a model based on maximum a posteriori policy optimization that can handle both numerical and categorical parameters is proposed. A case study compares the model to state–of–the–art traditional algorithms and shows that RL can find optima of similar quality while requiring significantly less time. These results are confirmed in a large-scale validation study on data sets from both production and other fields. Finally, multiple ways to improve the model are discussed.

Keywords: reinforcement learning, production process optimization, evolutionary algorithms, policy optimization, actor critic approach

Procedia PDF Downloads 90

Authors: Olaf Visker, Arnout De Vries, Lambert Schomaker

Abstract:

Predictive policing refers to the usage of analytical techniques to identify potential criminal activity. It has been widely implemented by various police departments. Being a relatively new area of research, there are, to the author’s knowledge, no absolute tried, and true methods and they still exhibit a variety of potential problems. One of those problems is closely related to the lack of understanding of how acting on these prediction influence crime itself. The goal of law enforcement is ultimately crime reduction. As such, a policy needs to be established that best facilitates this goal. This research aims to find such a policy by using adversarial agent-based modeling in combination with modern reinforcement learning techniques. It is presented here that a baseline model for both law enforcement and criminal agents and compare their performance to their respective reinforcement models. The experiments show that our smart law enforcement model is capable of reducing crime by making more deliberate choices regarding the locations of potential criminal activity. Furthermore, it is shown that the smart criminal model presents behavior consistent with popular crime theories and outperforms the baseline model in terms of crimes committed and time to capture. It does, however, still suffer from the difficulties of capturing long term rewards and learning how to handle multiple opposing goals.

Keywords: adversarial, agent based modelling, predictive policing, reinforcement learning

Procedia PDF Downloads 145

Authors: Wiem Zemzem, Moncef Tagina

Abstract:

In this work, we are addressing the problem of an autonomous mobile robot navigating in a large, unknown and dynamic environment using reinforcement learning abilities. This problem is principally related to the exploration/exploitation dilemma, especially the need to find a solution letting the robot detect the environmental change and also learn in order to adapt to the new environmental form without ignoring knowledge already acquired. Firstly, a new action selection strategy, called ε-greedy-MPA (the ε-greedy policy favoring the most promising actions) is proposed. Unlike existing exploration/exploitation policies (EEPs) such as ε-greedy and Boltzmann, the new EEP doesn’t only rely on the information of the actual state but also uses those of the eventual next states. Secondly, as the environment is large, an exploration favoring least recently visited states is added to the proposed EEP in order to accelerate learning. Finally, various simulations with ball-catching problem have been conducted to evaluate the ε-greedy-MPA policy. The results of simulated experiments show that combining this policy with the Qlearning method is more effective and efficient compared with the ε-greedy policy in stationary environments and the utility-based reinforcement learning approach in non stationary environments.

Keywords: autonomous mobile robot, exploration/ exploitation policy, large, dynamic environment, reinforcement learning

Procedia PDF Downloads 412

Authors: Romisaa Ali

Abstract:

This paper presents a comparison between twin-delayed Deep Deterministic Policy Gradient (TD3) and Soft Actor-Critic (SAC) reinforcement learning algorithms in the context of training robust navigation policies for Jackal robots. By leveraging an open-source framework and custom motion control environments, the study evaluates the performance, robustness, and transferability of the trained policies across a range of scenarios. The primary focus of the experiments is to assess the training process, the adaptability of the algorithms, and the robot’s ability to navigate in previously unseen environments. Moreover, the paper examines the influence of varying environmental complexities on the learning process and the generalization capabilities of the resulting policies. The results of this study aim to inform and guide the development of more efficient and practical reinforcement learning-based navigation policies for Jackal robots in real-world scenarios.

Keywords: Jackal robot environments, reinforcement learning, TD3, SAC, robust navigation, transferability, custom environment

Procedia PDF Downloads 93

Authors: Ahmed M. Eltohamy

Abstract:

In this paper, an experimental and numerical study was adopted to investigate the effect geogrid soil reinforcement pre-stressing on the pressure settlement relation of sand bed supporting a strip foundation. The studied parameters include foundation depth and pre-stress ratio for the cases of one and two pre-stressed reinforcement layers. The study reflected that pre-stressing of soil reinforcement resulted in a marked enhancement in reinforced bed soil stiffness compared to the reinforced soil without pre-stress. The best benefit of pre-stressing reinforcement was obtained as the overburden pressure and pre-straining ratio increase. Pre-stressing of double reinforcement topmost layers results in further enhancement of stress strain relation of bed soil.

Keywords: geogrid reinforcement, prestress, strip footing, bearing capacity

Procedia PDF Downloads 298

Authors: Arun Sanjel, Greg Speegle

Abstract:

Every second, massive amounts of data are generated, and Data Intensive Scalable Computing (DISC) frameworks have evolved into effective tools for analyzing such massive amounts of data. Since the underlying architecture of these distributed computing platforms is often new to users, building a DISC application can often be time-consuming and prone to errors. The automated conversion of a sequential program to a DISC program will consequently significantly improve productivity. However, synthesizing a user’s intended program from an input specification is complex, with several important applications, such as distributed program synthesizing and code refactoring. Existing works such as Tyro and Casper rely entirely on deductive synthesis techniques or similar program synthesis approaches. Our approach is to develop a data-driven synthesis technique to identify sequential components and translate them to equivalent distributed operations. We emphasize using reinforcement learning and unit testing as feedback mechanisms to achieve our objectives.

Keywords: program synthesis, distributed computing, reinforcement learning, unit testing, DISC

Procedia PDF Downloads 93

Authors: Vijaya Yuvaram Singh V M, Kameshwar Rao J V

Abstract:

The challenge with development of neural network based methods for medical is the availability of data. Anatomical landmark detection in the medical domain is a process to find points on the x-ray scan report of the patient. Most of the time this task is done manually by trained professionals as it requires precision and domain knowledge. Traditionally object detection based methods are used for landmark detection. Here, we utilize reinforcement learning and query based method to train a single agent capable of detecting multiple landmarks. A deep Q network agent is trained to detect single and multiple landmarks present on hip and shoulder from x-ray scan of a patient. Here a single agent is trained to find multiple landmark making it superior to having individual agents per landmark. For the initial study, five images of different patients are used as the environment and tested the agents performance on two unseen images.

Keywords: reinforcement learning, medical landmark detection, multi target detection, deep neural network

Procedia PDF Downloads 137

Authors: K. Passbakhsh

Abstract:

Determination of reinforcement forces is one of the most important and main discussions in designing retaining walls. By determining these forces we refrain from conservative planning. By numerically modeling the reinforced soil retaining walls under dynamic loading reinforcement forces can be calculated. In this study we try to approach the gained forces by pseudo-static method according to FHWA code and gained forces from numerical modeling by finite element method, by selecting seismic horizontal coefficient for different wall height. PLAXIS software was used for numerical analysis. Then the effect of reinforcement stiffness and soil type on reinforcement forces is examined.

Keywords: reinforced soil, PLAXIS, reinforcement forces, retaining walls

Procedia PDF Downloads 354

Authors: Rahul Paul, Kedar Nath Das

Abstract:

The field of machine learning (ML) is experiencing rapid development, resulting in a multitude of theoretical advancements and extensive practical implementations across various disciplines. The objective of ML is to facilitate the ability of machines to perform cognitive tasks by leveraging knowledge gained from prior experiences and effectively addressing complex problems, even in situations that deviate from previously encountered instances. Reinforcement Learning (RL) has emerged as a prominent subfield within ML and has gained considerable attention in recent times from researchers. This surge in interest can be attributed to the practical applications of RL, the increasing availability of data, and the rapid advancements in computing power. At the same time, optimization algorithms play a pivotal role in the field of ML and have attracted considerable interest from researchers. A multitude of proposals have been put forth to address optimization problems or improve optimization techniques within the domain of ML. The necessity of a thorough examination and implementation of optimization algorithms within the context of ML is of utmost importance in order to provide guidance for the advancement of research in both optimization and ML. This article provides a comprehensive overview of the application of metaheuristic evolutionary optimization algorithms in conjunction with RL to address a diverse range of scientific challenges. Furthermore, this article delves into the various challenges and unresolved issues pertaining to the optimization of RL models.

Keywords: machine learning, reinforcement learning, loss function, evolutionary optimization techniques

Procedia PDF Downloads 70

Authors: Y. Yongli, M. H. Aissa

Abstract:

The purpose of this paper aims for a geotechnical analysis based on experimental physical and mechanical characteristics of Miocene marl situated at Medea region in Algeria. More than 150 soil samples were taken in the investigation part of the North-South Highway which extends over than 53 km from Chiffa in the North to Berrouaghia in the South of Algeria. The analysis of data in terms of Atterberg limits, plasticity index, and clay content reflects an acceptable correlation justified by a high coefficient of regression which was compared with the previous works in the region. Finally, approximated equations that serve as a guideline for geotechnical design locally have been suggested.

Keywords: correlation, geotechnical properties, miocene marl, north-south highway

Procedia PDF Downloads 291

Authors: William Li

Abstract:

Lane-change path planning is a crucial and yet complex task in autonomous driving. The traditional path planning approach based on a system of carefully-crafted rules to cover various driving scenarios becomes unwieldy as more and more rules are added to deal with exceptions and corner cases. This paper proposes to divide the entire path planning to two stages. In the first stage the ego vehicle travels longitudinally in the source lane to reach a safe state. In the second stage the ego vehicle makes lateral lane-change maneuver to the target lane. The paper derives the safe state conditions based on lateral lane-change maneuver calculation to ensure collision free in the second stage. To determine the acceleration sequence that minimizes the time to reach a safe state in the first stage, the paper proposes three schemes, namely, kinetic model based optimization, deep reinforcement learning, and 5G vehicle-to-vehicle (V2V) communications. The paper investigates these schemes via simulation. The model-based optimization is sensitive to the model assumptions. The deep reinforcement learning is more flexible in handling scenarios beyond the model assumed by the optimization. The 5G V2V eliminates uncertainty in predicting future behaviors of surrounding vehicles by sharing driving intents and enabling cooperative driving.

Keywords: lane change, path planning, autonomous driving, deep reinforcement learning, 5G, V2V communications, connected vehicles

Procedia PDF Downloads 229

Authors: Bernardo Cortez, Florian Peter, Thomas Lausenhammer, Paulo Oliveira

Abstract:

Designing missiles’ autopilot controllers have been a complex task, given the extensive flight envelope and the nonlinear flight dynamics. A solution that can excel both in nominal performance and in robustness to uncertainties is still to be found. While Control Theory often debouches into parameters’ scheduling procedures, Reinforcement Learning has presented interesting results in ever more complex tasks, going from videogames to robotic tasks with continuous action domains. However, it still lacks clearer insights on how to find adequate reward functions and exploration strategies. To the best of our knowledge, this work is a pioneer in proposing Reinforcement Learning as a framework for flight control. In fact, it aims at training a model-free agent that can control the longitudinal non-linear flight dynamics of a missile, achieving the target performance and robustness to uncertainties. To that end, under TRPO’s methodology, the collected experience is augmented according to HER, stored in a replay buffer and sampled according to its significance. Not only does this work enhance the concept of prioritized experience replay into BPER, but it also reformulates HER, activating them both only when the training progress converges to suboptimal policies, in what is proposed as the SER methodology. The results show that it is possible both to achieve the target performance and to improve the agent’s robustness to uncertainties (with low damage on nominal performance) by further training it in non-nominal environments, therefore validating the proposed approach and encouraging future research in this field.

Keywords: Reinforcement Learning, flight control, HER, missile autopilot, TRPO

Procedia PDF Downloads 258

Authors: Albert Bou, Sebastian Dittert, Gianni de Fabritiis

Abstract:

Advancing reinforcement learning (RL) requires tools that are flexible enough to easily prototype new methods while avoiding impractically slow experimental turnaround times. To match the first requirement, the most popular RL libraries advocate for highly modular agent composability, which facilitates experimentation and development. To solve challenging environments within reasonable time frames, scaling RL to large sampling and computing resources has proved a successful strategy. However, this capability has been so far difficult to combine with modularity. In this work, we explore design choices to allow agent composability both at a local and distributed level of execution. We propose a versatile approach that allows the definition of RL agents at different scales through independent, reusable components. We demonstrate experimentally that our design choices allow us to reproduce classical benchmarks, explore multiple distributed architectures, and solve novel and complex environments while giving full control to the user in the agent definition and training scheme definition. We believe this work can provide useful insights to the next generation of RL libraries.

Keywords: deep reinforcement learning, Python, PyTorch, distributed training, modularity, library

Procedia PDF Downloads 78

Authors: Xiaosheng Chen, Jingjing Chen, Phil Reed, Dan Zhang

Abstract:

Conscious intention can be a promising point cut to grasp consciousness and orient voluntary action. The current study adopted a random ratio (RR), yoked random interval (RI) reinforcement learning schedule instead of the previous highly repeatable and single decision point paradigms, aimed to induce voluntary action with the conscious intention that evolves from the interaction between short-range-intention and long-range-intention. Readiness potential (RP) -like-EEG amplitude and inter-trial-EEG variability decreased significantly prior to voluntary action compared to cued action for inter-trial-EEG variability, mainly featured during the earlier stage of neural activities. Notably, (RP) -like-EEG amplitudes decreased significantly prior to higher RI-reward rates responses in which participants formed a higher plane of conscious intention. The present study suggests the possible contribution of conscious intention-based processes to the neural activities from the earlier stage prior to voluntary action on reinforcement leanring schedule.

Keywords: Reinforcement leaning schedule, voluntary action, EEG, conscious intention, readiness potential

Procedia PDF Downloads 69

Authors: Yuling Cui, Danhao Deng, Chaowei Wang, Weidong Wang

Abstract:

In recent years, unmanned aerial vehicles (UAVs) have been widely used in wireless communication, attracting more and more attention from researchers. UAVs can not only serve as a relay for auxiliary communication but also serve as an aerial base station for ground users (GUs). However, limited energy means that they cannot work all the time and cover a limited range of services. In this paper, we investigate 2D UAV trajectory design and power allocation in order to maximize the UAV's service time and downlink throughput. Based on deep reinforcement learning, we propose a depth deterministic strategy gradient algorithm for trajectory design and power distribution (TDPA-DDPG) to solve the energy-efficient and communication service quality problem. The simulation results show that TDPA-DDPG can extend the service time of UAV as much as possible, improve the communication service quality, and realize the maximization of downlink throughput, which is significantly improved compared with existing methods.

Keywords: UAV trajectory design, power allocation, energy efficient, downlink throughput, deep reinforcement learning, DDPG

Procedia PDF Downloads 143

Authors: Fares Jnaid, Riyad Aboutaha

Abstract:

In this paper, a nonlinear Finite Element Analysis (FEA) was carried out using ANSYS software to build a model able of predicting the behavior of Reinforced Concrete (RC) beams with unbonded reinforcement. The FEA model was compared to existing experimental data by other researchers. The existing experimental data consisted of 16 beams that varied from structurally sound beams to beams with unbonded reinforcement with different unbonded lengths and reinforcement ratios. The model was able to predict the ultimate flexural strength, load-deflection curve, and crack pattern of concrete beams with unbonded reinforcement. It was concluded that when the when the unbonded length is less than 45% of the span, there will be no decrease in the ultimate flexural strength due to the loss of bond between the steel reinforcement and the surrounding concrete regardless of the reinforcement ratio. Moreover, when the reinforcement ratio is relatively low, there will be no decrease in ultimate flexural strength regardless of the length of unbond.

Keywords: FEA, ANSYS, unbond, strain

Procedia PDF Downloads 250

Authors: Liping Qiu, Xiaofeng Zhang

Abstract:

In its long history, precious tomb murals suffered from various diseases due to natural and man-made destruction. The key to how to protect tomb murals is how to strengthen and protect the tomb murals. In order to maximize the life of the tomb murals, the artistic, historic, and scientific values of the tomb murals can be continued. In this paper, four kinds of traditional reinforcement materials (silicone acrylic lotion, pure acrylic lotion, polyvinyl acetate lotion, and B72) are selected to reinforce the ground support layer of tomb murals, and the reinforcement effect of each reinforcement material on the ground support layer of murals is compared and analyzed, and the best protection material is obtained.

Keywords: mural, destruction cycle, reinforcement material, disease

Procedia PDF Downloads 119

Authors: Szymon Rutkowski, Tomasz Korbak

Abstract:

We present a reinforcement-learning-based method of training neural machine translation models without parallel corpora. The standard encoder-decoder approach to machine translation suffers from two problems we aim to address. First, it needs parallel corpora, which are scarce, especially for low-resource languages. Second, it lacks psychological plausibility of learning procedure: learning a foreign language is about learning to communicate useful information, not merely learning to transduce from one language’s 'encoding' to another. We instead pose the problem of learning to translate as learning a policy in a communication game between two agents: the translator and the classifier. The classifier is trained beforehand on a natural language inference task (determining the entailment relation between a premise and a hypothesis) in the target language. The translator produces a sequence of actions that correspond to generating translations of both the hypothesis and premise, which are then passed to the classifier. The translator is rewarded for classifier’s performance on determining entailment between sentences translated by the translator to disciple’s native language. Translator’s performance thus reflects its ability to communicate useful information to the classifier. In effect, we train a machine translation model without the need for parallel corpora altogether. While similar reinforcement learning formulations for zero-shot translation were proposed before, there is a number of improvements we introduce. While prior research aimed at grounding the translation task in the physical world by evaluating agents on an image captioning task, we found that using a linguistic task is more sample-efficient. Natural language inference (also known as recognizing textual entailment) captures semantic properties of sentence pairs that are poorly correlated with semantic similarity, thus enforcing basic understanding of the role played by compositionality. It has been shown that models trained recognizing textual entailment produce high-quality general-purpose sentence embeddings transferrable to other tasks. We use stanford natural language inference (SNLI) dataset as well as its analogous datasets for French (XNLI) and Polish (CDSCorpus). Textual entailment corpora can be obtained relatively easily for any language, which makes our approach more extensible to low-resource languages than traditional approaches based on parallel corpora. We evaluated a number of reinforcement learning algorithms (including policy gradients and actor-critic) to solve the problem of translator’s policy optimization and found that our attempts yield some promising improvements over previous approaches to reinforcement-learning based zero-shot machine translation.

Keywords: agent-based language learning, low-resource translation, natural language inference, neural machine translation, reinforcement learning

Procedia PDF Downloads 124

Authors: Akshay Paranjape, Nils Plettenberg, Robert Schmitt

Abstract:

Traditional optimization methods like evolutionary algorithms are widely used in production processes to find an optimal or near-optimal solution of control parameters based on the simulated environment space of a process. These algorithms are computationally intensive and therefore do not provide the opportunity for real-time optimization. This paper utilizes the Deep Reinforcement Learning (DRL) framework to find an optimal or near-optimal solution for control parameters. A model based on maximum a posteriori policy optimization (Hybrid-MPO) that can handle both numerical and categorical parameters is used as a benchmark for comparison. A comparative study shows that DRL can find optimal solutions of similar quality as compared to evolutionary algorithms while requiring significantly less time making them preferable for real-time optimization. The results are confirmed in a large-scale validation study on datasets from production and other fields. A trained XGBoost model is used as a surrogate for process simulation. Finally, multiple ways to improve the model are discussed.

Keywords: reinforcement learning, evolutionary algorithms, production process optimization, real-time optimization, hybrid-MPO

Procedia PDF Downloads 106

Authors: Almutasim Billa A. Alanazi, Hal S. Tharp

Abstract:

Safe performance and efficient energy consumption are essential factors for designing a control system. This paper presents a reinforcement learning (RL) model that can be applied to control applications to improve safety and reduce energy consumption. As hardware constraints and environmental disturbances are imprecise and unpredictable, conventional control methods may not always be effective in optimizing control designs. However, RL has demonstrated its value in several artificial intelligence (AI) applications, especially in the field of control systems. The proposed model intelligently monitors a system's success by observing the rewards from the environment, with positive rewards counting as a success when the controlled reference is within the desired operating zone. Thus, the model can determine whether the system is safe to continue operating based on the designer/user specifications, which can be adjusted as needed. Additionally, the controller keeps track of energy consumption to improve energy efficiency by enabling the idle mode when the controlled reference is within the desired operating zone, thus reducing the system energy consumption during the controlling operation. Water temperature control for a hydroponic system is taken as a case study for the RL model, adjusting the variance of disturbances to show the model’s robustness and efficiency. On average, the model showed safety improvement by up to 15% and energy efficiency improvements by 35%- 40% compared to a traditional RL model.

Keywords: control system, hydroponics, machine learning, reinforcement learning

Procedia PDF Downloads 173

Authors: Tong He, Long Chen, Irag Mantegh, Wen-Fang Xie

Abstract:

This paper proposes an image-based obstacle avoidance and tracking target identification strategy in GPS-degraded or GPS-denied environment for an Unmanned Aerial Vehicle (UAV). The traditional force algorithm for obstacle avoidance could produce local minima area, in which UAV cannot get away obstacle effectively. In order to eliminate it, an artificial potential approach based on harmonic potential is proposed to guide the UAV to avoid the obstacle by using the vision system. And image-based visual servoing scheme (IBVS) has been adopted to implement the proposed obstacle avoidance approach. In IBVS, the pixel accuracy is a key factor to realize the obstacle avoidance. In this paper, the deep reinforcement learning framework has been applied by reducing pixel errors through constant interaction between the environment and the agent. In addition, the combination of OpenTLD and Tensorflow based on neural network is used to identify the type of tracking target. Numerical simulation in Matlab and ROS GAZEBO show the satisfactory result in target identification and obstacle avoidance.

Keywords: image-based visual servoing, obstacle avoidance, tracking target identification, deep reinforcement learning, artificial potential approach, neural network

Procedia PDF Downloads 134

Authors: Xiaohua Zou, Yongxin Su

Abstract:

The reasonable operation of heating, ventilation, and air conditioning (HVAC) is of great significance in improving the security, stability, and economy of power system operation. However, the uncertainty of the operating environment, thermal comfort varies by users and rapid decision-making pose challenges for HVAC demand response optimization. In this regard, this paper proposes a reinforcement learning-based method for HVAC demand response optimization considering few-shot personalized thermal comfort (PTC). First, an HVAC DR optimization framework based on few-shot PTC model and DRL is designed, in which the output of few-shot PTC model is regarded as the input of DRL. Then, a few-shot PTC model that distinguishes between awake and asleep states is established, which has excellent engineering usability. Next, based on soft actor criticism, an HVAC DR optimization algorithm considering the user’s PTC is designed to deal with uncertainty and make decisions rapidly. Experiment results show that the proposed method can efficiently obtain use’s PTC temperature, reduce energy cost while ensuring user’s PTC, and achieve rapid decision-making under uncertainty.

Keywords: HVAC, few-shot personalized thermal comfort, deep reinforcement learning, demand response

Procedia PDF Downloads 75