Complexity analysis method and system for automatic driving test scene

Abstract

The invention discloses an automatic driving test scene complexity analysis method, which relates to the technical field of unmanned driving, and comprises the following steps: carrying out frame-by-frame information extraction on a target scene fragment; establishing a single-frame scene data complexity analysis architecture; for each frame of scene data in the target scene segment, constructing a fuzzy judgment matrix of a corresponding subordinate hierarchy for each superior hierarchy in the frame of scene data; performing defuzzification processing on the fuzzy judgment matrix to obtain a weight value of each evaluation index relative to the target layer; performing standardization processing on each evaluation index to obtain the complexity of the frame of scene data; and carrying out weighted average on the complexity of all frame data to obtain the complexity of the target scene. According to the method, the expert experience and the scene complexity analysis and evaluation scheme based on data driving are combined, the expert experience is fully utilized and converted into quantifiable information, the problem of deviation of the expert experience is solved, and the accuracy of analysis of the complexity of the automatic driving test scene is higher.

Description

technical field [0001] The invention relates to the technical field of unmanned driving, in particular to a method and system for analyzing the complexity of an automatic driving test scene. Background technique [0002] At present, domestic and foreign related enterprises or scientific research institutions are focusing on the algorithm iterative optimization of each module of autonomous driving, but the corresponding test and evaluation of autonomous driving vehicles is still in its infancy. There is still a lack of systematic theoretical research and support in construction, which makes the evaluation of test scenarios weaker in systematization, and lacks a reasonable and effective system to evaluate the complexity of various typical autonomous driving test scenarios, resulting in problems in the design and selection of scenarios. There is a certain degree of redundancy, and it is also difficult to distinguish the performance differences of the current automatic driving s...

AI Technical Summary

Problems solved by technology

[0004] For the experience-driven analysis scheme, it is mainly for each traffic participant in the scene data and the relevant environmental information at that time, and the corresponding experts will conduct a comprehensive evaluation and scoring based on their own experience, so as to obtain an evaluation of the scene As a result, the problem with this scheme is that there are differences in the experience and subjective feelings of various experts, and it cannot be effectively quantitatively evaluated.

The other is quantitative evaluation based on data, such as the calculation method of road traffic environment complexity based on the gravity model, using information entropy theory to establish a traffic environment complexity model, and realizing the specific quantification of the static and dynamic complexity of the road traffic environment, but this This type relies heavily on the scene database. In the case of insufficient scene data, the accuracy of scene complexity evaluation will be affected. At the same time, some indicators that are not easy to quantify cannot be well included in the complexity analysis of this category of scenes. , and did not make full use of expert experience

Images