Mingheng Zhang

Automatic collision avoidance decision making for vessels is a critical challenge in the development of autonomous ships and has become a central point of research in the maritime safety domain. Effective and systematic collision avoidance strategies significantly reduce the risk of vessel collisions, ensuring safe navigation. This study develops a multi-vessel automatic collision avoidance decision-making method based on deep reinforcement learning (DRL) and establishes a vessel behavior decision model. When designing the reward function for continuous action spaces, the criteria of the “Convention on the International Regulations for Preventing Collisions at Sea” (COLREGs) were adhered to, taking into account the vessel’s collision risk under various encounter situations, real-world navigation practices, and navigational complexities. Furthermore, to enable the algorithm to precisely differentiate between collision avoidance and the navigation resumption phase in varied vessel encounter situations, this paper incorporated “collision avoidance decision making” and “course recovery decision making” as state parameters in the state set design, from which the respective objective functions were defined. To further enhance the algorithm’s performance, techniques such as behavior cloning, residual networks, and CPU-GPU dual-core parallel processing modules were integrated. Through simulation experiments in the enhanced Imazu training environment, the practicality of the method, taking into account the effects of wind and ocean currents, was corroborated. The results demonstrate that the proposed algorithm can perform effective collision avoidance decision making in a range of vessel encounter situations, indicating its efficiency and robust generalization capabilities.

As urban traffic pollution continues to increase, there is an urgent need to build traffic emission monitoring and forecasting system for the urban traffic construction. The traffic emission monitoring and forecasting system's core is the prediction of traffic emission's evolution. And the traffic flow prediction on the urban road network contributes greatly to the prediction of traffic emission's evolution. Due to the complex non-Euclidean topological structure of traffic networks and dynamic heterogeneous spatial-temporal correlations of traffic conditions, it is difficult to obtain satisfactory prediction results with less computation cost. To figure these issues out, a novel deep learning traffic flow forecasting framework is proposed in this paper, termed as Ensemble Attention based Graph Time Convolutional Networks (EAGTCN). More specifically, each component of our model contains two major blocks: (1) the global spatial patterns are captured by the spatial blocks which are fused by the Graph Convolution Network (GCN) and spatial ensemble attention layer; (2) the temporal patterns are captured by the temporal blocks which are composed by the Time Convolution Net (TCN) and temporal ensemble attention layers. Experiments on two real-world datasets demonstrate that our model obtains more accurate prediction results than the state-of-the-art baselines at less computation expense especially in the long-term prediction situation.

Reliability, efficiency and generalization are basic evaluation criteria for a vehicle automated driving system. This paper proposes an automated driving...