Autonomous Driving Decision-Making With Sequential Strategy Space Release
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving technologies face challenges in efficiently making intelligent driving decisions due to high computational requirements, especially in complex scenarios where multiple obstacles are involved, leading to increased computing power demands that are often beyond the capabilities of existing hardware.
Innovation Solution
The method involves releasing multiple strategy spaces in a sequential manner, spanning dimensions such as longitudinal, lateral, and temporal sampling, to determine a strategy feasible region for the ego vehicle and game objects, thereby minimizing the number of strategy spaces released and reducing the computational burden while ensuring decision-making precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple strategy spaces are released to ensure decision-making precision, then decision-making accuracy is improved, but computing power consumption increases
Solution Approach 1:
The patent segments the strategy space release process into multiple sequential stages (first release, second release, third release) with increasing granularity. Each stage releases only the necessary portion of strategy spaces required at that level, avoiding the need to compute all strategy spaces simultaneously. This segmented approach maintains decision-making precision while reducing peak computing power consumption.
Solution Approach 2:
The patent performs preliminary action by releasing coarse-grained strategy spaces first (longitudinal and lateral dimensions) before releasing finer-grained strategy spaces (temporal dimensions). This preliminary release of essential strategy spaces establishes a foundation for decision-making without requiring all detailed strategy spaces to be computed at once, thereby reducing overall computing power requirements.
2Reliability
If comprehensive strategy spaces are searched to ensure decision accuracy, then decision-making reliability is improved, but computational complexity increases
Solution Approach 1:
The patent divides the comprehensive strategy space search into segmented phases: first releasing longitudinal sampling strategy spaces, then lateral sampling strategy spaces, and finally temporal sampling strategy spaces. Each phase processes a subset of strategy spaces with specific dimensions, reducing the computational complexity of each individual processing step while maintaining overall decision-making reliability through the cumulative effect of all phases.
Solution Approach 2:
The patent applies dynamics by adaptively adjusting the release of strategy spaces based on decision-making needs. The system dynamically releases strategy spaces in three progressive stages, where each stage releases additional strategy spaces only when necessary. This dynamic approach ensures reliability by releasing comprehensive strategy spaces when needed while reducing computational complexity by avoiding premature release of all strategy spaces.
Data Source
AI summary
This application relates to intelligent driving technologies, and provides an intelligent driving decision-making method, including: first, obtaining a game object of an ego vehicle; then from a plurality of strategy spaces of the ego vehicle and the game object, performing a plurality of times of release of the plurality of strategy spaces; and determining a strategy feasible region of the ego vehicle and the game object based on each released strategy space, and determining a traveling decision-making result of the ego vehicle based on the strategy feasible region. The decision-making result is an executable behavior action of the ego vehicle. As described above, by releasing the strategy spaces for a plurality of times, while decision-making precision is ensured, the decision-making result may be obtained when as fewer strategy spaces are released as possible. This reduces a computing amount and lowers a requirement for hardware computing power.