Automated vehicle’s behavior decision making using deep reinforcement learning and high-fidelity simulation environment
تصمیم گیری خودکار وسیله نقلیه با استفاده از یادگیری تقویتی عمیق و محیط شبیه سازی با وفاداری بالا-2019
Automated vehicles (AVs) are deemed to be the key element for the intelligent transportation system in the future. Many studies have been made to improve AVs’ ability of environment recognition and vehicle control, while the attention paid to decision making is not enough and the existing decision algorithms are very preliminary. Therefore, a framework of the decisionmaking training and learning is put forward in this paper. It consists of two parts: the deep reinforcement learning (DRL) training program and the high-fidelity virtual simulation environment. Then the basic microscopic behavior, car-following (CF), is trained within this framework. In addition, theoretical analysis and experiments were conducted to evaluate the proposed reward functions for accelerating training using DRL. The results show that on the premise of driving comfort, the efficiency of the trained AV increases 7.9% and 3.8% respectively compared to the classical adaptive cruise control models, intelligent driver model and constant-time headway policy. Moreover, on a more complex three-lane section, we trained an integrated model combining both CF and lane-changing behavior, with the average speed further growing 2.4%. It indicates that our framework is effective for AV’s decision-making learning.
Keywords: Automated vehicle | Decision making | Deep reinforcement learning | Reward function
(Deep) Reinforcement learning for electric power system control and related problems: A short review and perspectives
(عمیق) یادگیری تقویتی برای کنترل سیستم برق و مشکلات مرتبط با آن: یک مرور کوتاه و چشم اندازها-2019
This paper reviews existing works on (deep) reinforcement learning considerations in electric power sys- tem control. The works are reviewed as they relate to electric power system operating states (normal, preventive, emergency, restorative) and control levels (local, household, microgrid, subsystem, wide-area). Due attention is paid to the control-related problems considerations (cyber-security, big data analysis, short-term load forecast, and composite load modelling). Observations from reviewed literature are drawn and perspectives discussed. In order to make the text compact and as easy as possible to read, the focus is only on the works published (or “in press”) in journals and books while conference publications are not included. Exceptions are several work available in open repositories likely to become journal pub- lications in near future. Hopefully this paper could serve as a good source of information for all those interested in solving similar problems.
Keywords: Electric power system | Reinforcement learning | Deep reinforcement learning | Control | Control-related problems
Intelligent fault diagnosis for rotating machinery using deep Q-network based health state classification: A deep reinforcement learning approach
تشخیص خطای هوشمند برای ماشین آلات در حال چرخش با استفاده از طبقه بندی حالت سلامت مبتنی بر شبکه Q عمقی: یک روش یادگیری تقویتی عمیق-2019
Fault diagnosis methods for rotating machinery have always been a hot research topic, and artificial intelligencebased approaches have attracted increasing attention from both researchers and engineers. Among those related studies and methods, artificial neural networks, especially deep learning-based methods, are widely used to extract fault features or classify fault features obtained by other signal processing techniques. Although such methods could solve the fault diagnosis problems of rotating machinery, there are still two deficiencies. (1) Unable to establish direct linear or non-linear mapping between raw data and the corresponding fault modes, the performance of such fault diagnosis methods highly depends on the quality of the extracted features. (2) The optimization of neural network architecture and parameters, especially for deep neural networks, requires considerable manual modification and expert experience, which limits the applicability and generalization of such methods. As a remarkable breakthrough in artificial intelligence, AlphaGo, a representative achievement of deep reinforcement learning, provides inspiration and direction for the aforementioned shortcomings. Combining the advantages of deep learning and reinforcement learning, deep reinforcement learning is able to build an end-to-end fault diagnosis architecture that can directly map raw fault data to the corresponding fault modes. Thus, based on deep reinforcement learning, a novel intelligent diagnosis method is proposed that is able to overcome the shortcomings of the aforementioned diagnosis methods. Validation tests of the proposed method are carried out using datasets of two types of rotating machinery, rolling bearings and hydraulic pumps, which contain a large number of measured raw vibration signals under different health states and working conditions. The diagnosis results show that the proposed method is able to obtain intelligent fault diagnosis agents that can mine the relationships between the raw vibration signals and fault modes autonomously and effectively. Considering that the learning process of the proposed method depends only on the replayed memories of the agent and the overall rewards, which represent much weaker feedback than that obtained by the supervised learning-based method, the proposed method is promising in establishing a general fault diagnosis architecture for rotating machinery.
Keywords: Fault diagnosis | Rotating machinery | Deep reinforcement learning | Deep Q-network
Decentralized network level adaptive signal control by multi-agent deep reinforcement learning
کنترل سیگنال تطبیقی سطح شبکه غیر متمرکز با یادگیری تقویت عمیق چند عاملی-2019
Adaptive traffic signal control systems are deployed to accommodate real-time traffic conditions. Yet travel demand and behavior of the individual vehicles might be overseen by their model-based control algorithms and aggregated input data. Recent development of artificial intelligence, especially the success of deep learning, makes it possible to utilize information of individual vehicles to control the traffic signals. Several pioneering studies developed modelfree control algorithms using deep reinforcement learning. However, those studies are limited to isolated intersections and their effectiveness was only evaluated in ideal simulated traffic conditions by hypothetical benchmarks. To fill the gap, this study proposes a network-level decentralized adaptive signal control algorithmusing one of the famous deep reinforcement methods, double dueling deep Q network in the multi-agent reinforcement learning framework. The proposed algorithm was evaluated by the real-world coordinated actuated signals in a simulated suburban traffic corridor which emulates the real-field traffic condition. The evaluation results showed that the proposed deepreinforcement- learning-based algorithm outperforms the benchmark. It is able to reduce 10.27% of the travel time and 46.46% of the total delay.
Keywords: Deep reinforcement learning | Multi-agent reinforcement learning | Adaptive signal control
Deep reinforcement learning-based controller for path following of an unmanned surface vehicle
کنترلر مبتنی بر یادگیری تقویتی عمیق برای پیگیری مسیر یک وسیله نقلیه سطحی بدون سرنشین-2019
In this paper, a deep reinforcement learning (DRL)-based controller for path following of an unmanned surface vehicle (USV) is proposed. The proposed controller can self-develop a vehicle’s path following capability by interacting with the nearby environment. A deep deterministic policy gradient (DDPG) algorithm, which is an actor-critic-based reinforcement learning algorithm, was adapted to capture the USV’s experience during the path-following trials. A Markov decision process model, which includes the state, action, and reward formulation, specially designed for the USV path-following problem is suggested. The control policy was trained with repeated trials of path-following simulation. The proposed method’s path-following and self-learning capabilities were validated through USV simulation and a free-running test of the full-scale USV.
Keywords: Deep reinforcement learning | Path following | Unmanned surface vehicle | Learning-based control | Artificial intelligence
Deep reinforcement learning with its application for lung cancer detection in medical Internet of Things
یادگیری تقویتی عمیق با کاربرد آن برای تشخیص سرطان ریه در اینترنت اشیاء پزشکی -2019
Recently, deep reinforcement learning has achieved great success by integrating deep learning models into reinforcement learning algorithms in various applications such as computer games and robots. Specially, it is promising for computer-aided diagnosis and treatment to combine deep reinforcement learning with medical big data generated and collected from medical Internet of Things. In this paper, we focus on the potential of the deep reinforcement learning for lung cancer detection as many people are suffering from the lung tumor and about 1.8 million patients died from lung cancer in 2018. Early detection and diagnosis of lung tumor can significantly improve the treatment effect and prolong survival. In this work, we present several representative deep reinforcement learning models that are potential to use for lung cancer detection. Furthermore, we summarize the common types of lung cancer and the main characteristics of each type. Finally, we point out the open challenges and possible future research directions of applying deep reinforcement learning to lung cancer detection, which is expected to promote the evolution of smart medicine with medical Internet of Things.
Keywords: Smart medicine | Medical Internet of Things | Deep reinforcement learning | Lung cancer
Evaluating smart grid renewable energy accommodation capability with uncertain generation using deep reinforcement learning
ارزیابی قابلیت اسکان انرژی تجدیدپذیر شبکه هوشمند با تولید نامعین با استفاده از یادگیری تقویتی عمیق-2019
Due to environment-friendliness, renewable energy like solar power and wind power is more and more introduced to energy systems all over the world. Simultaneously, high penetrations of wind and solar generation also have brought severe curtailment of wind and solar. How to alleviate curtailment of wind and solar is a crucial problem in evaluating accommodation capability of renewable energy, which reflects the extent of utilization of renewable energy and economic benefits. The uncertainty of renewable energy brings challenges to precisely describe renewable generation, which leads to difficulty in designing effective mechanisms for accommodation capability of renewable energy. Existing work suffers from high computation overhead from frequently updated data, and low precision of describing renewable energy, which leads to less effective policies for renewable energy accommodation and underestimated accommodation capability. To make the most of renewable energy, an algorithm AccCap-DRL based on deep reinforcement learning is proposed. AccCap-DRL partitions a distribution into segments by time intervals, employs WGAN to describe distributions of renewable energy data, and employs DDPG to obtain approximate policies for renewable energy accommodation in different scenarios. Simulation results from real power generation and users’ demand data show high effectiveness of the proposed algorithm, and high efficiency of evaluating accommodation capability
Keywords: Accommodation capability | Deep reinforcement learning | Uncertain renewable energy description
Deep learning for decision making and the optimization of socially responsible investments and portfolio
یادگیری عمیق برای تصمیم گیری و بهینه سازی سرمایه گذاری ها و نمونه کارها با مسئولیت اجتماعی-2019
A socially responsible investment portfolio takes into consideration the environmental, social and governance aspects of companies. It has become an emerging topic for both financial investors and researchers recently. Traditional investment and portfolio theories, which are used for the optimization of financial investment portfolios, are inadequate for decision-making and the construction of an optimized socially responsible investment portfolio. In response to this problem, we introduced a Deep Responsible Investment Portfolio (DRIP) model that contains a Multivariate Bidirectional Long Short-Term Memory neural network, to predict stock returns for the construction of a socially responsible investment portfolio. The deep reinforcement learning technique was adapted to retrain neural networks and rebalance the portfolio periodically. Our empirical data revealed that the DRIP framework could achieve competitive financial performance and better social impact compared to traditional portfolio models, sustainable indexes and funds.
Keywords: Socially responsible investment | Portfolio optimization | Multivariate analytics | Deep reinforcement learning | Decision support systems
Deep Reinforcement Learning based Patch Selection for Illuminant Estimation
انتخاب پچ مبتنی بر یادگیری تقویتی عمیق برای برآورد روشنایی-2019
Previous deep learning based approaches to illuminant estimation either resized the raw image to lower resolution or randomly cropped image patches for the deep learning model. However, such practices would inevitably lead to information loss or the selection of noisy patches that would aect estimation accuracy. In this paper, we regard patch selection in neural network based illuminant estimation as a controlling problem of selecting image patches that could help remove noisy patches and improve estimation accuracy. To achieve this, we construct a selection network (SeNet) to learn a patch selection policy. Based on data statistics and the learning progression state of the deep illuminant estimation network (DeNet), the SeNet decides which training patches should be input to the DeNet, which in turn gives feedback to the SeNet for it to update its selection policy. To achieve such interactive and intelligent learning, we utilize a reinforcement learning approach termed policy gradient to optimize the SeNet. We show that the proposed learning strategy can enhance the illuminant estimation accuracy, speed up the convergence and improve the stability of the training process of DeNet. We evaluate our method on two public datasets and demonstrate our method outperforms state-of-the-art approaches.
Keywords: Color constancy | reinforcement learning | patch selection
Energy management for a power-split hybrid electric bus via deep reinforcement learning with terrain information
مدیریت انرژی برای یک اتوبوس برقی هیبریدی با تقسیم قدرت از طریق یادگیری تقویتی عمیق با اطلاعات زمین-2019
Due to the high mileage and heavy load capabilities of hybrid commercial vehicles, energy management becomes crucial in improving their fuel economy. In this paper, terrain information is systematically integrated into the energy management strategy for a power-split hybrid electric bus based on a deep reinforcement learning approach: the deep deterministic policy gradient algorithm. Specially, this energy management method is improved and capable of searching optimal energy management strategies in a discrete-continuous hybrid action space, which, in this work, consists of two continuous actions for the engine and four discrete actions for powertrain mode selections. Additionally, a Critic network with dueling architecture and a pre-training stage ahead of the reinforcement learning process are combined for efficient strategy learning with the adopted algorithm. Assuming the current terrain information was available to the controller, the deep reinforcement learning based energy management strategy is trained and tested on different driving cycles and simulated terrains. Simulation results of the trained strategy show that reasonable energy allocation schemes and mode switching rules are learned simultaneously. Its fuel economy gap with the baseline strategy using dynamic programming is narrowed down to nearly 6.4% while reducing the times of engine starts by around 76%. Further comparisons also indicate approximately 2% promotion in fuel economy is contributed by the incorporation of terrain information in this learning-based energy management. The main contribution of this study is to explore the inclusion of terrain information in a learning-based energy management method that can deal with large hybrid action spaces.
Keywords: Energy management strategy | Deep reinforcement learning | Discrete-continuous hybrid action space | Power-split hybrid electric bus | Terrain information