A rapid evaluation method and system for autonomous capability of unmanned system decision-making strategy

Abstract

The invention discloses a method and system for quickly evaluating the autonomous ability of an unmanned system's decision-making strategy. The steps include: 1) constructing a dangerous scene knowledge base, and realizing the corresponding dangerous scene in a simulation environment; 2) being tested according to the input 3) Iteratively test the measured object starting from the dangerous environment of low risk level: first, for each risk level j, select N dangerous scenes belonging to the risk level j as the test scene Simulation test, if the pass rate of the i-th test scenario i is P i ≥Y i Then mark the i-th test scenario test as passed, then let P i '=1, otherwise P i '=0; then calculate the risk value of level j if R j ≥Z j , then mark the hazard level j as passed; 4) Generate a test report according to the test results of each hazard level.

Description

technical field [0001] The invention relates to a rapid evaluation method and system for autonomous ability of an unmanned system decision-making strategy, and belongs to the technical field of computer software. Background technique [0002] Autonomy means that the unmanned system has the ability to perceive, observe, analyze, communicate, plan, make decisions and act, and complete the tasks assigned to it by humans through human-computer interaction, and then express the state and quality of self-management of the unmanned system. A fast and effective evaluation method can not only ensure that the lack of autonomy of the unmanned system is found, but also effectively promote the development of its autonomy. [0003] The core of intelligent unmanned system technology is the decision-making strategy, which involves a lot of information interaction in the development process, such as the status information of the intelligent unmanned system itself, surrounding environment inf...

AI Technical Summary

Problems solved by technology

In this technical solution, a complexity model of driving conditions is proposed, but because it is only for driving vehicles and requires the cooperation of the driver, the test period is long and the safety factor of the driver is required to be high; and the dangerous driving work considered The situation is limited to natural driving vehicles, so the influencing factors of the hazard scene complexity model are relatively one-sided

[0006] The paper "Research on Prediction of Random Motion State of Front Vehicle Based on Cluster Analysis of Hazardous Scenes" uses knowledge-based and data-based methods to screen dangerous scenarios based on five elements: environment information, own vehicle information, target information, traffic information, and road information , but the usage scenario is based on natural driving, so the consideration of dangerous elements and dangerous scenarios has certain limitations

Images