Search

Traffic congestion has become a major problem in this rapidly growing world. Everyone operating a vehicle, as well as the traffic police in charge of managing the traffic, finds it difficult to become stuck in heavy traffic. For this a set, predetermined timing for traffic flow for each direction at the junction is utilized by traditional traffic light controllers. However, the concept of a fixed time traffic signal controller does not work well in places with uneven traffic. A dynamic traffic control system is therefore required, which regulates the traffic signals in accordance with the volume of traffic. This paper proposes a model that uses reinforcement learning (RL) along with deep neural networks (DNN) to manage discretions (signal status) for an environment with the help of Simulation of Urban MObility (SUMO). A simulation of real-world environment consisting a network of Four-way crossroad junction that contains 4 arriving lanes and 4 exiting lanes is used to train the agent. The main objective of this research study is to construct a model that can independently determine the best course of action and aims to provide better traffic management that will decrease the average waiting time, cause lower congestion, and provide a smooth flow of traffic. [ABSTRACT FROM AUTHOR]

Healthcare systems are burdened by mobility impairments resulting from aging and neurological conditions. One of the recent advances in robotics is lower limb assistive/rehabilitative devices that can make independent living possible. Nonetheless, some limitations need to be addressed before robotics can be used in home‐based applications. This paper describes the current state of the art in intelligent motion planning and control of lower limb assistive devices, which have addressed some of these challenges. Adaptable central pattern generators and the divergent component of motion are introduced as methods for personalized motion planning based on physical human–robot interaction (pHRI). Uncertainty analysis for neural networks is introduced to increase safety in motion planning based on pHRI. For the case that a user cannot apply physical interaction, a reinforcement‐learning‐based approach is introduced to switch between different modes of walking based on the user's input via a push button embedded in a walker. Moreover, a smart walker is introduced as a device that can be synchronized with the lower limb exoskeleton to assist users with their daily activities. Also, a roadmap for future steps that can make lower limb assistive/rehabilitative devices a good fit for home use is introduced. [ABSTRACT FROM AUTHOR]

In advanced transportation-management systems, variable speed limits are a crucial application. Deep reinforcement learning methods have been shown to have superior performance in many applications, as they are an effective approach to learning environment dynamics for decision-making and control. However, they face two significant difficulties in traffic-control applications: reward engineering with delayed reward and brittle convergence properties with gradient descent. To address these challenges, evolutionary strategies are well suited as a class of black-box optimization techniques inspired by natural evolution. Additionally, the traditional deep reinforcement learning framework struggles to handle the delayed reward setting. This paper proposes a novel approach using covariance matrix adaptation evolution strategy (CMA-ES), a gradient-free global optimization method, to handle the task of multi-lane differential variable speed limit control. The proposed method uses a deep-learning-based method to dynamically learn optimal and distinct speed limits among lanes. The parameters of the neural network are sampled using a multivariate normal distribution, and the dependencies between the variables are represented by a covariance matrix that is optimized dynamically by CMA-ES based on the freeway's throughput. The proposed approach is tested on a freeway with simulated recurrent bottlenecks, and the experimental results show that it outperforms deep reinforcement learning-based approaches, traditional evolutionary search methods, and the no-control scenario. Our proposed method demonstrates a 23% improvement in average travel time and an average of a 4% improvement in CO, HC, and NOx emission.Furthermore, the proposed method produces explainable speed limits and has desirable generalization power. [ABSTRACT FROM AUTHOR]

• A graph neural network is proposed for learning multi-satellite scheduling policy. • It analyzes the differential impact of various constraints in scheduling. • It models the multi-satellite scheduling problem as a heterogeneous graph. • It designs a variant transformer to learn various relations within graph. • It designs a two-stage model to learn a multi-satellite scheduling policy. In this paper, we investigate the multi-satellite scheduling (MSS) problem, a combinatorial optimization problem (COP) that aims to enhance satellite utilization. Numerous scholars have conducted in-depth studies on this matter. Handcrafted policies are commonly employed in various algorithms. Despite the use of graphical representations, designing policies manually becomes increasingly difficult, particularly as the number of satellites and tasks grows. Consequently, we propose a novel approach for automatically summarizing graph structures and learning policies for MSS problem. The proposed approach is developed based on graph neural network, Transformer, and sequence to sequence framework. First, MSS problem is modeled as a heterogeneous graph following various constraints. Next, the encoder, which is based on GNN and Transformer, is designed to summarize crucial information from the aforementioned graph. Finally, a two-stage decoder is proposed taking into account dynamic information in stepwise solution generation. Since obtaining optimal solutions in practical applications is challenging, we train the model using reinforcement learning. The results of the experiment demonstrate that our proposed method performs effectively in various scenarios. The quality of the solutions generated through this method transcends that of conventional construction methods. Additionally, the method generates solutions quickly, which are comparable in quality to those produced by meta-heuristic algorithms. [ABSTRACT FROM AUTHOR]

Hydropower is the most prevalent source of renewable energy production worldwide. As the global demand for robust and ecologically sustainable energy production increases, developing and enhancing the current energy production processes is essential. In the past decade, machine learning has contributed significantly to various fields, and hydropower is no exception. All three horizons of hydropower models could benefit from machine learning: short-term, medium-term, and long-term. Currently, dynamic programming is used in the majority of hydropower scheduling models. In this paper, we review the present state of the hydropower scheduling problem as well as the development of machine learning as a type of optimization problem and prediction tool. To the best of our knowledge, this is the first survey article that provides a comprehensive overview of machine learning and artificial intelligence applications in the hydroelectric power industry for scheduling, optimization, and prediction. [ABSTRACT FROM AUTHOR]

To prevent the proliferation of space debris and stabilize the space environment, active debris removal (ADR) has increasingly gained public concern. Considering the complexity of space operations and the viability of ADR missions, it would be necessary to schedule the ADR process in order to remove as much debris as possible. This paper presents an active debris removal mission planning problem, devoted to generate an optimal debris removal plan to guide the mission process. According to the problem characteristics, a two-layer time-dependent traveling salesman problem(TSP) mathematical model is established, involving the debris removal sequence planning and the transfer trajectory planning. Subsequently, two main novel methods based on machine learning are proposed for the ADR mission planning problem, including a deep neural networks(DNN)-based estimation method for approximating the optimal velocity increments of perturbed multiple-impulse rendezvous and an reinforcement learning(RL)-based method for optimizing the sequence of debris removal and rendezvous time. Experimental results of different simulation scenarios have verified the effectiveness and superiority of the proposed method, indicating the good performance for solving the active debris removal mission planning problem. [ABSTRACT FROM AUTHOR]

The weak interpretability of the deep reinforcement learning (DRL) model becomes a serious impediment to the application of DRL agents in certain areas requiring high reliability. To interpret the behavior of a DRL agent, researchers use saliency maps to discover important parts of the agent's observation that influence its decision. However, the representations of saliency maps still cannot explicitly present the cause and effect between an agent's actions and its observations. In this paper, we analyze the inference procedure with respect to the DRL architecture and propose embedding interpretable intermediate representations for an agent's policy, the intermediate representations that are compressed and abstracted for explanation. We utilize a conceptual embedding technique to regulate the latent representation space of the deep models that can produce interpretable causal factors aligned with human concepts. Furthermore, the information loss of intermediate representation is analyzed to define the model performance upper bound and to measure the model performance degeneration. Experiments validate the effectiveness of the proposed method and the relationship between the observation information and an agent's performance upper bound. [ABSTRACT FROM AUTHOR]

As a promising information theory, reinforcement learning has gained much attention. This paper researches a wind-storage cooperative decision-making strategy based on dueling double deep Q-network (D3QN). Firstly, a new wind-storage cooperative model is proposed. Besides wind farms, energy storage systems, and external power grids, demand response loads are also considered, including residential price response loads and thermostatically controlled loads (TCLs). Then, a novel wind-storage cooperative decision-making mechanism is proposed, which combines the direct control of TCLs with the indirect control of residential price response loads. In addition, a kind of deep reinforcement learning algorithm called D3QN is utilized to solve the wind-storage cooperative decision-making problem. Finally, the numerical results verify the effectiveness of D3QN for optimizing the decision-making strategy of a wind-storage cooperation system. [ABSTRACT FROM AUTHOR]

In the resource-constrained embedded systems, the designing of efficient deep neural networks is a challenging process, due to diversity in the artificial intelligence applications. The quantization in deep neural networks superiorly diminishes the storage and computational time by reducing the bit-width of networks encoding. In order to highlight the problem of accuracy loss, the quantization levels are automatically discovered using Policy Feedback-based Reinforcement Learning Method (PF-RELEQ). In this paper, the Proximal Policy Optimization with Policy Feedback (PPO-PF) technique is proposed to determine the best design decisions by choosing the optimum hyper-parameters. In order to enhance the sensitivity of the value function to the change of policy and to improve the accuracy of value estimation at the early learning stage, a policy update method is devised based on the clipped discount factor. In addition, specifically the loss functions of policy satisfy the unbiased estimation of the trust region. The proposed PF-RELEQ effectively balances quality and speed compared to other deep learning methods like ResNet-1202, ResNet-32, ResNet-110, GoogLeNet and AlexNet. The experimental analysis showed that PF-RELEQ achieved 20% computational work-load reduction compared to the existing deep learning methods on ImageNet, CIFAR-10, CIFAR-100 and tomato leaf disease datasets and achieved approximately 2% of improvisation in the validation accuracy. Additionally, the PF-RELEQ needs only 0.55 Graphics Processing Unit on an NVIDIA GTX-1080Ti to develop DNNs that delivers better accuracy improvement with fewer cycle counts for image classification. [ABSTRACT FROM AUTHOR]

Image caption technology aims to convert visual features of images, extracted by computers, into meaningful semantic information. Therefore, the computers can generate text descriptions that resemble human perception, enabling tasks such as image classification, retrieval, and analysis. In recent years, the performance of image caption has been significantly enhanced with the introduction of encoder-decoder architecture in machine translation and the utilization of deep neural networks. However, several challenges still persist in this domain. Therefore, this paper proposes a novel method to address the issue of visual information loss and non-dynamic adjustment of input images during decoding. We introduce a guided decoding network that establishes a connection between the encoding and decoding parts. Through this connection, encoding information can provide guidance to the decoding process, facilitating automatic adjustment of the decoding information. In addition, Dense Convolutional Network (DenseNet) and Multiple Instance Learning (MIL) are adopted in the image encoder, and Nested Long Short-Term Memory (NLSTM) is utilized as the decoder to enhance the extraction and parsing capability of image information during the encoding and decoding process. In order to further improve the performance of our image caption model, this study incorporates an attention mechanism to focus details and constructs a double-layer decoding structure, which facilitates the enhancement of the model in terms of providing more detailed descriptions and enriched semantic information. Furthermore, the Deep Reinforcement Learning (DRL) method is employed to train the model by directly optimizing the identical set of evaluation indexes, which solves the problem of inconsistent training and evaluation standards. Finally, the model is trained and tested on MS COCO and Flickr 30 k datasets, and the results show that the model has improved compared with commonly used models in the evaluation indicators such as BLEU, METEOR and CIDEr. [ABSTRACT FROM AUTHOR]

The positional information of objects is crucial to enable robots to perform grasping and pushing manipulations in clutter. To effectively perform grasping and pushing manipulations, robots need to perceive the position information of objects, including the coordinates and spatial relationship between objects (e.g., proximity, adjacency). The authors propose an end‐to‐end position‐aware deep Q‐learning framework to achieve efficient collaborative pushing and grasping in clutter. Specifically, a pair of conjugate pushing and grasping attention modules are proposed to capture the position information of objects and generate high‐quality affordance maps of operating positions with features of pushing and grasping operations. In addition, the authors propose an object isolation metric and clutter metric based on instance segmentation to measure the spatial relationships between objects in cluttered environments. To further enhance the perception capacity of position information of the objects, the authors associate the change in the object isolation metric and clutter metric in cluttered environment before and after performing the action with reward function. A series of experiments are carried out in simulation and real‐world which indicate that the method improves sample efficiency, task completion rate, grasping success rate and action efficiency compared to state‐of‐the‐art end‐to‐end methods. Noted that the authors' system can be robustly applied to real‐world use and extended to novel objects. Supplementary material is available at https://youtu.be/NhG\_k5v3NnM}{https://youtu.be/NhG\_k5v3NnM. [ABSTRACT FROM AUTHOR]

This paper addresses how physical knowledge can improve machine learning in process control. A data-driven tracking control framework using physics-informed neural networks (PINNs) and deep reinforcement learning (DRL) is proposed for dynamical systems, which is particularly important when iterations or repetitions of data collection experiments are limited or even not allowed. An indirect control approach is followed to accomplish this. First, a new process model is identified using PINNs trained from a model representing the experimental data collected from the plant. Then, the DRL-based control agent utilizes the identified PINN model to deliver an offline policy. We evaluate our framework for data-driven tracking control of nonisothermal CSTR, considering measurement noise and stochastic dynamics for closed-loop control. Our approach performed comparably to an NMPC without requiring a model to predict the process dynamics. • Physical knowledge was implemented to improve machine learning. • A indirect data-driven control approach was developed. • Physics-informed neural networks and deep reinforcement learning are combined. • The developed framework was tested against nonlinear model predictive control. [ABSTRACT FROM AUTHOR]

A segmentation model of the ultrasound (US) images of breast tumors based on virtual agents trained using reinforcement learning (RL) is proposed. The agents, living in the edge map, are able to avoid false boundaries, connect broken parts, and finally, accurately delineate the contour of the tumor. The agents move similarly to robots navigating in the unknown environment with the goal of maximizing the rewards. The individual agent does not know the goal of the entire population. However, since the robots communicate, the model is able to understand the global information and fit the irregular boundaries of complicated objects. Combining the RL with a neural network makes it possible to automatically learn and select the local features. In particular, the agents handle the edge leaks and artifacts typical for the US images. The proposed model outperforms 13 state-of-the-art algorithms, including selected deep learning models and their modifications. [ABSTRACT FROM AUTHOR]

The demand for autonomous motion control of unmanned aerial vehicles in air combat is boosted as taking the initiative in combat appears more and more crucial. Unmanned aerial vehicles inability to manoeuvre autonomously during air combat that features highly dynamic and uncertain manoeuvres of the enemy; however, limits their combat capabilities, which proves to be very challenging. To meet the challenge, this article proposes an autonomous manoeuvre decision model using an expert actor‐based soft actor critic algorithm that reconstructs empirical replay buffer with expert experience. Specifically, the algorithm uses a small amount of expert experience to increase the diversity of the samples, which can largely improve the exploration and utilisation efficiency of deep reinforcement learning. And to simulate the complex battlefield environment, a one‐to‐one air combat model is established and the concept of missile's attack region is introduced. The model enables the one‐to‐one air combat to be simulated under different initial battlefield situations. Simulation results show that the expert actor‐based soft actor critic algorithm can find the most favourable policy for unmanned aerial vehicles to defeat the opponent faster, and converge more quickly, compared with the soft actor critic algorithm. [ABSTRACT FROM AUTHOR]

The payoff of an agent depends on both the environment and the actions of other agents. Thus, the ability to model and predict the strategies and behaviors of other agents in an interactive decision-making scenario is one of the core functionalities in intelligent systems. State-of-the-art methods for opponent modeling mainly use an explicit model of opponents' actions, preferences, targets, etc., that the primary agent uses to make decisions. It is more important for an agent to increase its payoff than to accurately predict opponents' behavior. Therefore, we propose a framework synchronizing the opponent modeling and decision making of the primary agent by incorporating opponent modeling into reinforcement learning. For interactive decisions, the payoff depends not only on the behavioral characteristics of the opponent but also the current state. However, confounding the two obscures the effects of state and action, which then cannot be accurately encoded. To this end, state evaluation is separated from action evaluation in our model. The experimental results from two game environments, a simulated soccer game and a real game called quiz bowl, show that the introduction of opponent modeling can effectively improve decision payoffs. In addition, the proposed framework for opponent modeling outperforms benchmark models. [ABSTRACT FROM AUTHOR]

It is well known that unmodeled dynamics and uncertainties can deteriorate the performance of classical controllers. To resolve this problem, there is growing popularity in using the capabilities of Artificial Intelligence (AI) algorithms, especially Reinforcement Learning (RL) in power systems, because it is a promising adaptive model-free control strategy that can take optimal decisions in unknown environments (dynamics). For this reason, in this paper, two state-of-the-art RL agents, namely Deep Q-Network (DQN) and Deep Deterministic Policy Gradient (DDPG), are used for voltage control of a DC–DC buck converter, and their performance is reported compared with other classical controllers such as Model Predictive Control (MPC) and Sliding Mode Control (SMC). The DQN agent directly controls the power switches of converters. In other words, based on the current condition of the converter, the agent decides whether or not to close the power switches. On the other hand, the DDPG agent and the other mentioned traditional controllers manipulate the duty cycle of a Pulse Width Modulation (PWM) signal to adjust the output voltage of the converter at desired setpoints. According to experimental results, both RL agents outperform the classical controllers in terms of transient response error and robustness against uncertainties. Also, with regard to computational costs and learning rate among RL-based controllers, the DQN agent can learn more from a single interaction with fewer computations because of its simpler structure and direct control of the switches of the converter. Additionally, one of the most important advantages of the RL-based controllers is that they can be applied to various configurations of DC–DC​ converters like buck, boost, and buck-boost converters, provided that it is retrained for the new environments. Finally, the number of transitions in the semiconductor switches of the converter reduces appreciably by using the DQN agent, which certainly prolongs their longevity. [ABSTRACT FROM AUTHOR]

Reinforcement learning-based controllers for safety-critical applications, such as autonomous driving, are typically trained in simulation, where a vehicle model is provided during the learning process. However, an inaccurate parameterization of the vehicle model used for training heavily influences the performance of the reinforcement learning agent during execution. This inaccuracy is either caused by changes due to environmental influences or by falsely estimated vehicle parameters. In this work, we present our approach of combining dynamics randomization with reinforcement learning to overcome this issue for a path-following control task of an autonomous and over-actuated robotic vehicle. We train three independent agents, where each agent experiences randomization for a different vehicle dynamics parameter, i.e., the mass, the yaw inertia, and the road-tire friction. We randomize the parameters uniformly within predefined ranges to enable the agents to learn an equally robust control behavior for all possible parameter values. Finally, in a simulation study, we compare the performance of the agents trained with dynamics randomization to the performance of an agent trained with the nominal parameter values. Simulation results demonstrate that the former agents obtain a higher level of robustness against model uncertainties and varying environmental conditions than the latter agent trained with nominal vehicle parameter values. [ABSTRACT FROM AUTHOR]

Reinforcement learning (RL) empowers the agent to learn robotic manipulation skills autonomously. Compared with traditional single-goal RL, semantic-goal-conditioned RL expands the agent capacity to accomplish multiple semantic manipulation instructions. However, due to sparsely distributed semantic goals and sparse-reward agent-environment interactions, the hard exploration problem arises and impedes the agent training process. In traditional RL, curiosity-motivated exploration shows effectiveness in solving the hard exploration problem. However, in semantic-goal-conditioned RL, the performance of previous curiosity-motivated methods deteriorates, which we propose is because of their two defects: uncontrollability and distraction. To solve these defects, we propose a conservative curiosity-motivated method named mutual information motivation with hybrid policy mechanism (MIHM). MIHM mainly contributes two innovations: the decoupled-mutual-information- based intrinsic motivation, which prevents the agent from being motivated to explore dangerous states by uncontrollable curiosity; the precisely trained and automatically switched hybrid policy mechanism, which eliminates the distraction from the curiosity-motivated policy and achieves the optimal utilization of exploration and exploitation. Compared with four state-of-the-art curiosity-motivated methods in the sparse-reward robotic manipulation task with 35 valid semantic goals, including stacks of 2 or 3 objects and pyramids, our MIHM shows the fastest learning speed. Moreover, MIHM achieves the highest 0.9 total success rate, which is up to 0.6 in other methods. Throughout all the baseline methods, our MIHM is the only one that achieves to stack three objects. [ABSTRACT FROM AUTHOR]

Deep Reinforcement Learning (DRL) algorithms have been widely studied for sequential decision-making problems, and substantial progress has been achieved, especially in autonomous robotic skill learning. However, it is always difficult to deploy DRL methods in practical safety-critical robot systems, since the training and deployment environment gap always exists, and this issue would become increasingly crucial due to the ever-changing environment. Aiming at efficiently robotic skill transferring in a dynamic environment, we present a meta-reinforcement learning algorithm based on a variational information bottleneck. More specifically, during the meta-training stage, the variational information bottleneck first has been applied to infer the complete basic tasks for the whole task space, then the maximum entropy regularized reinforcement learning framework has been used to learn the basic skills consistent with that of basic tasks. Once the training stage is completed, all of the tasks in the task space can be obtained by a nonlinear combination of the basic tasks, thus, the according skills to accomplish the tasks can also be obtained by some way of a combination of the basic skills. Empirical results on several highly nonlinear, high-dimensional robotic locomotion tasks show that the proposed variational information bottleneck regularized deep reinforcement learning algorithm can improve sample efficiency by 200–5000 times on new tasks. Furthermore, the proposed algorithm achieves substantial asymptotic performance improvement. The results indicate that the proposed meta-reinforcement learning framework makes a significant step forward to deploy the DRL-based algorithm to practical robot systems. [ABSTRACT FROM AUTHOR]