Driver control system and method for relieving traffic jam, medium, equipment and application

Abstract

The invention belongs to the technical field of auxiliary driving in automatic driving, and discloses a driver control system and method for relieving traffic jam, a medium, equipment and application, and the method comprises the steps: employing the setting of centralized learning but decentralized execution, making a decision for each target vehicle node at each moment, achieving the same given target for all nodes, and enabling the target vehicle node to be a target vehicle node; therefore, the traffic jam problem is solved. Modeling is carried out on communication and information propagation between nodes by adopting a graph neural network GNN, a decision processor adopts Deep Q learning, and formed decision information is issued to drivers in each environment in the form of suggestion instructions. The method has the advantage of scene traversal depth and self-learning performance, all working conditions can be covered more easily through a big data system, a machine can extract environment characteristics and decision attributes by themselves, and system optimization iteration is facilitated; the model is perfected through data training, and the model accuracy is improved along with data completeness.

Description

technical field [0001] The invention belongs to the technical field of driving assistance in automatic driving, and in particular relates to a driver control system, method, medium, device and application for alleviating traffic jams. Background technique [0002] At present, with the increasing demand for transportation and travel, the traffic pressure on the urban road network is increasing, and the problem of traffic congestion is becoming more and more serious. In the urban road network, the occurrence of congestion is not independent. When traffic congestion occurs at a certain node in the traffic network, if it cannot be evacuated in time, the congestion will spread throughout the network, and then lead to large-scale network congestion. Described as successive congestion. The traditional solution is to increase infrastructure construction, but if no corresponding planning and management strategies are adopted, it is difficult to achieve the desired effect of congesti...

AI Technical Summary

Problems solved by technology

[0019] (1) The traditional solution is to increase infrastructure construction, but if no corresponding planning and management strategies are adopted, it is difficult to achieve the desired effect of congestion relief

[0020] (2) The existing technology has never actually solved the problem of traffic congestion, and stopped at the prediction and proposal of the problem in the real traffic system, so its function is not perfect

[0021] (3) The existing technology will further increase the complexity of the system for the traffic jams that inevitably occur in large cities during peak traffic periods, and the uncertainty of driver behavior that is common during traffic jams, and the above patented methods are not applicable

[0022] The difficulty in solving the above problems and defects is: in traffic scenes with a large number of interactive processes such as intersections and fork roads, serious traffic congestion often occurs. Relatively global cognition, so driven by the maximization of self-interest utility, self-interested driving strategies are often adopted, but this process often further aggravates the traffic congestion problem;

[0023] In the traffic scene of the above-mentioned complex game, the relative relationship between each individual vehicle node is intricate, including relative position relationship, relative motion relationship, competition relationship and cooperation relationship, etc. In the process of decision-making and planning, the functional core is still Mathematical solver, so all the above information needs to be considered and quantified at the same time, but the game relationship such as competition and cooperation has dynamic uncertainty, and is different from traditional data types, so it is difficult to effectively quantify

Images