Unmanned vehicle multi-objective decision-making control method, device and computer-readable storage medium

Abstract

The invention discloses a multi-objective decision-making control method and device for an unmanned vehicle, and a computer-readable storage medium. Vehicles; if there is a vehicle in the area of ​​interest, determine whether the vehicle has a tendency to approach and whether the vehicle has a line; according to whether there is a target vehicle in the area of ​​interest, whether the vehicle in the area of ​​interest has a tendency to approach and whether there is a line situation, judge the danger level of the vehicle, and output the scene label; make decisions on the expected behavior of the vehicle based on the danger level. Based on the existing unmanned driving application field, the present invention increases the understanding of multi-target scenarios, and activates corresponding control modules according to different scenarios, so that the behavior of the unmanned vehicle is close to that of a skilled driver, and the ride experience of the occupants is improved. and vehicle safety.

Description

technical field [0001] The invention belongs to the technical field of unmanned vehicle environment cognition, and in particular relates to a multi-objective decision-making control method, device and computer-readable storage medium of an unmanned vehicle. Background technique [0002] In the current engineering application field, the decision-making control of unmanned vehicles is generally based on single-target following ACC, which adaptively adjusts the state of the vehicle only according to the state of a vehicle in the lane. The limitation of this algorithm is that it can only be applied to lower levels of automatic driving and simpler working conditions, such as highway conditions with relatively small traffic flow. In order to meet higher levels of automatic driving, such as automatic driving on congested expressways, the traffic flow is relatively large, the driving environment is relatively complex, and there are many uncertainties. If only following the target ve...

AI Technical Summary

Problems solved by technology

The hazard level assessment based on the test data model has many limitations: first, obtaining test data, extracting characteristic parameters, clustering of special parameters and establishing a data model require a large number of tests, which require a lot of manpower and material resources; Second, there is a large uncertainty when extracting feature data, and it is difficult to ensure that the required data can be extracted at the same time in different scenarios; third, using this method to map scenarios and models requires the use of Machine learning, existing machine learning methods, such as support vector machines and decision trees, currently have no commercially viable solutions

The target vehicle trajectory prediction method based on the long and short time domain also has many limitations: first, the short time domain prediction is generally based on the vehicle kinematics model, which is only suitable for vehicle trajectory prediction in a short period of time, and its prediction error increases with the prediction time Second, long-term predictions are generally based on behavioral cognition models, such as dynamic Bayesian network models. Dynamic Bayesian networks involve prior probabilities, which are difficult to obtain, so the application prospects are The short-term prediction accuracy of this model is poor

The comprehensive time domain prediction model developed on the basis of long time domain prediction and short time domain prediction has the advantages of both long time domain accuracy prediction and short time domain prediction. However, the application of this method is much more difficult than long time domain prediction and short time domain prediction. Time-domain prediction is high, practical application prospects are limited

Images