Search

In many real-life scenarios, multiple agents necessitate cooperation to accomplish tasks. Benefiting from the significant success of deep learning, many single-agent deep reinforcement learning algorithms have been extended to multi-agent scenarios. Overestimation in value estimation of Q-learning is a significant issue that has been studied comprehensively in the single-agent domains, but rarely in multi-agent reinforcement learning. In this paper, we first demonstrate that Q-learning-based multi-agent reinforcement learning (MARL) methods generally have notably serious overestimation issues, which cannot be alleviated by current methods. To tackle this problem, we introduce the double critic networks structure and the delayed policy update to Q-learning-based multi-agent MARL methods, which reduce the overestimation and enhance the quality of policy updating. To demonstrate the versatility of our proposed method, we select several Q-learning based MARL methods and evaluate them on several multi-agent tasks on the multi-agent particle environment and SMAC. Experimental results demonstrate that the proposed method can avoid the overestimation problem and significantly improve performance. Besides, application in the Traffic Signal Control verifies the feasibility of applying the proposed method in real-world scenarios. [ABSTRACT FROM AUTHOR]

Accurate assessment of road conditions can effectively alleviate traffic congestion and guide people's travel plans, traffic control decisions of transportation departments, and formulation of traffic-related laws and regulations. This paper proposes a quantum particle swarm optimization (QPSO) and adaptive hybrid exponential smoothing with residual correction (AHES-RC) for the nonlinearity and randomness of traffic flow. In the proposed algorithm, the single–double-exponential smoothing method is mixed with adaptive weights to form AHES, a new method of mixing weights according to real-time traffic trend changes. After the residual correction of AHES is calculated by the extreme learning machine algorithm, the parameters of AHES-RC are optimized using QPSO to improve the prediction accuracy further. This paper comprehensively compares the QPSO-AHES-RC algorithm with other benchmark models through testing on 26 real-world data sets. The results show that the proposed algorithm can adaptive forecasting under various traffic conditions, yielding the best forecasting performance in terms of various forecast error metrics. Compared with advanced machine learning algorithms such as XGBoost and CatBoost, the mean RMSE and mean MAPE have been improved by over 20% on average. In addition, it is revealed that the real-time dynamic capture of traffic flow can effectively improve prediction accuracy. [ABSTRACT FROM AUTHOR]

Driverless automatically guided vehicles are becoming a new trend in transportation envisioned in a smart city environment. Automatically guided vehicles depend on environmental data for executing target-oriented navigation and movements. The traffic knowledge can be used by automatically guided vehicles by measuring their intensity and delivery date in the green light. Therefore, considering sensor data's significance in this guided vehicle environment, this article presents the Internet of Things (IoT) assisted automated guided vehicle (IoT-AGV) scheme. IoT-AGV scheme discusses the implementation of an autonomous vehicle traffic control program at crossings. The traffic knowledge operator is ready to retrieve details from the sensor utilizing quick communication through a mobile network to measure optimum travel time and transmit the information collected to automated cars. In an automated guided vehicle (AGV) scheme, the automotive-into-loop model is introduced to eliminate delays. Error indicator integrates various hurdle imbalance statistics calculated from the sensor based on the road slide. A deep neural network focuses on a minimal collection of the most critical features, eliminating the fitting problem. This aids determine the error producing sensor data and alleviate them from fusion to enhance the target's accuracy. The experimental results show that the suggested system has been validated by the images from BIT vehicle datasets and enhances the accuracy ratio of 96.22% to measure optimum travel time and transmit the information collected to automated guided vehicles. [ABSTRACT FROM AUTHOR]