Autonomous Vehicle Behavior Control via Multi-Stage Grading
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in effectively planning and controlling their behavior, particularly in mid-range and short-range movements, as they need to consider interactions with other vehicles and unexpected objects while maintaining safety and efficiency.
Innovation Solution
A system and method that uses sensor inputs to generate and grade multiple possible movements using autonomous driving techniques like rule-based models and machine learning tree-regression models, selecting the best movement based on scoring parameters such as speed, distance to objects, and presence of pedestrians or stop signs, and then regrading these movements to account for predicted actions of other vehicles, ensuring optimal path planning and reactive movements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple possible planned movements are generated and graded using autonomous driving techniques, then the decision-making accuracy is improved, but the computational complexity and processing time increase
Solution Approach 1:
The behavior planning is divided into multiple independent stages: generating multiple possible planned movements, grading each movement using autonomous driving techniques, determining predicted movements of other vehicles, and selecting the best movement. This segmentation allows each stage to be processed separately, improving decision-making accuracy while managing computational complexity through modular processing.
Solution Approach 2:
The system performs preliminary grading of multiple possible planned movements before final selection. By pre-evaluating and scoring different movement options using rule-based models and machine learning tree-regression models, the system prepares decision candidates in advance, reducing the computational burden during real-time execution and improving overall decision accuracy.
2Reliability
If the system considers interactions with other vehicles and regrades movements based on predicted actions, then the safety is improved, but the processing time increases
Solution Approach 1:
The system determines predicted movements of other vehicles in advance based on the selected planned movement. By pre-predicting how other vehicles will react to the autonomous vehicle's intended actions, the system can regrade movements more efficiently and make safer decisions without excessive processing delays.
Solution Approach 2:
The system regrades the planned movements after considering the predicted reactions of other vehicles. This feedback loop allows the autonomous vehicle to adjust its decisions based on how other road users are likely to respond, improving safety by accounting for interactive dynamics while using efficient algorithms to minimize processing time.
3Adaptability or versatility
If multiple factors such as speed, distance to objects, and presence of pedestrians are considered in grading, then the comprehensiveness of decision-making is improved, but the complexity of evaluation increases
Solution Approach 1:
The grading process is segmented into multiple independent evaluation dimensions: speed assessment, distance to objects, presence of pedestrians, presence of stop signs, and other factors. Each dimension is evaluated separately using specific autonomous driving techniques, allowing comprehensive decision-making while managing evaluation complexity through structured modular assessment.
Solution Approach 2:
The system uses universal autonomous driving techniques including rule-based models and machine learning tree-regression models that can evaluate multiple factors simultaneously. These multi-functional evaluation methods assess various parameters (speed, distance, pedestrians, signs) through a unified framework, improving comprehensiveness while reducing the complexity of implementing separate evaluation systems for each factor.
Data Source
AI summary
A control system of the autonomous vehicle may generate multiple possible behavior control movements based on the driving goal and the assessment of the vehicle environment. In doing so, the method and system selects one of the best behavior control, among the multiple possible movements, and the selection is based on the quantitative grading of its driving behavior.