This paper proposes a convolutional neural network (CNN)-based method that learns traffic as images and predicts large-scale, network-wide traffic speed with a high accuracy. Spatiotemporal traffic dynamics are converted to images describing the time and space relations of traffic flow via a two-dimensional time-space matrix. A CNN is applied to the image following two consecutive steps: abstract traffic feature extraction and network-wide traffic speed prediction. The effectiveness of the proposed method is evaluated by taking two real-world transportation networks, the second ring road and north-east transportation network in Beijing, as examples, and comparing the method with four prevailing algorithms, namely, ordinary least squares, k-nearest neighbors, artificial neural network, and random forest, and three deep learning architectures, namely, stacked autoencoder, recurrent neural network, and long-short-term memory network. The results show that the proposed method outperforms other algorithms by an average accuracy improvement of 42.91% within an acceptable execution time. The CNN can train the model in a reasonable time and, thus, is suitable for large-scale transportation networks.

In the upcoming decades, connected vehicles will join regular vehicles to appear on roads, and the characteristics of traffic flow will be changed accordingly. To model the heterogeneous traffic mixing regular and connected vehicles, a generic car-following framework is first proposed in this paper. A linear stability condition is theoretically derived, which indicates that the stability of the heterogeneous traffic is closely related to the penetration rate and the spatial distribution of connected vehicles. The generic car-following framework is applied by taking the Intelligent Driver Model as an example, and it is shown that connected vehicles can obviously enhance the stability of traffic flow and improve traffic efficiency in particular when traffic is in congestion. Moreover, a driver assistance strategy based on distributed feedback control is developed for connected vehicles, and the simulation results show that the proposed driver assistance strategy performs satisfactorily in stabilizing traffic as well as improving traffic efficiency.

For autonomous vehicles driving on roads, future trajectories of surrounding traffic agents (e.g., vehicles, bicycles, pedestrians) are essential information. The prediction of future trajectories is challenging as the motion of traffic agents is constantly affected by spatial-temporal interactions from agents and road infrastructure. To take those interactions into account, this study proposes a Graph Attention Transformer (Gatformer) in which a traffic scene is represented by a sparse graph. To maintain the spatial and temporal information of traffic agents in a traffic scene, Convolutional Neural Networks (CNNs) are utilized to extract spatial features and a position encoder is proposed to encode the spatial features and the corresponding temporal features. Based on the encoded features, a Graph Attention Network (GAT) block is employed to model the agent-agent and agent-infrastructure interactions with the help of attention mechanisms. Finally, a Transformer network is introduced to predict trajectories for multiple agents simultaneously. Experiments are conducted over the Lyft dataset and state-of-the-art methods are introduced for comparison. The results show that the proposed Gatformer could make more accurate predictions while requiring less inference time than its counterparts.

One of the most challenging problems in human-machine co-work is the gap between human intention and the machine's understanding and execution. Large Language Models (LLMs) have been showing superior abilities in solving such issue. In this paper, we design a universal framework that embeds LLMs as a vehicle “Co-Pilot” of driving, which can accomplish specific driving tasks with human intention satisfied based on the information provided. Meanwhile, a utilization workflow is defined to handle the interaction between humans and vehicles, and memory mechanism is introduced to organize the information involved in the tasks. Expert Oriented Black-Box tuning is proposed to improve the performance of the Co-Pilot without finetuning or training the LLMs. In the experiment, the Co-Pilot is applied to two different tasks, i.e., path tracking control and trajectory planning. The Co-Pilot adjusts vehicle operating conditions by selecting a proper controller or planning a certain trajectory to fit human intentions. Simulation tests are conducted to evaluate the performance and generality of the proposed module. The results show that the Co-Pilot can accomplish most of the tasks based on only natural language processing, although it is not flawless. Finally, a discussion about human-machine hybrid intelligence and further applications of LLMs in autonomous driving is made. We believe that such a framework has promising potential in further applications in the field of automous vehicles.

To mitigate traffic oscillations that usually sustainably propagate upstream, this paper proposes a jam-absorption driving (JAD) strategy in the framework of Newell's car-following theory. The basic idea of the JAD strategy is to guide a vehicle to move slowly before being captured by an oscillation and terminate the slow movement when the vehicle would start to leave the jam if no such slow movement was implemented. To practically implement the idea, a two-step method is proposed to estimate the time-space ending point of the strategy, and a proper vehicle is selected to implement the JAD strategy based on a given expected absorbing speed and current traffic conditions. To test the JAD strategy, two simulated traffic scenarios are constructed based on a realistic data-driven car-following model. The first scenario, which only reproduces one oscillation, directly shows the effectiveness of the JAD idea in preventing wave propagation and capacity drop. The second scenario, which contains a series of traffic oscillations induced by the rubbernecking behavior, validates the proposed JAD strategy in more complicated and realistic conditions. It is indicated that the JAD strategy is able to absorb traffic oscillations; thus, the side effects incurred by the oscillations could be subsequently mitigated. The significance of this paper is to provide us a new idea to mitigate traffic oscillations, i.e., the JAD strategy.

Imagine in the future that autonomous vehicles coordinated and guided by signal-free autonomous intersections are able to pass through the intersections immediately after the vehicles in the conflicting direction leave. Meanwhile, with the coordination of the autonomous intersections, autonomous vehicles on any approaching lane are able to turn onto any downstream lane, expecting that driving in such a road system, autonomous vehicles can reach their destinations without any on-road lane changes, and high traffic efficiency will be achieved as well as great traffic safety. To draw the picture in detail, this paper designs a signal-free autonomous intersection with all-direction turn lanes (ADTL) under the environment of autonomous vehicles, and proposes a conflict-avoidance-based approach to coordinate all approaching vehicles in different directions. Communicating with the approaching autonomous vehicles and utilizing the approach, the autonomous ADTL intersection is able to coordinate the approaching vehicles in all directions and guide them to safely and efficiently pass through the intersection. Two simulation scenarios are conducted in a road network with an isolated intersection composed of four three-lane arms. One scenario validates the collision-free design of the system, and the other shows that the designed ADTL intersection outperforms the conventional signal controlled intersection in terms of traffic efficiency, and is potentially better than the autonomous intersection with specific-direction turn lanes. The autonomous ADTL intersection can be an important basis for designing a future autonomous urban road traffic system.