Autonomous Vehicle Interaction Phases for Yielding Decision Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving methods fail to accurately predict the behavior of interacting vehicles, leading to increased collision risk and poor ride experience due to inconsistent decision-making with actual vehicle actions, especially in scenarios where one vehicle yields to another.
Innovation Solution
The method divides the interaction process into multiple phases, using perception information to dynamically adjust decision solutions based on the interaction status, allowing for continuous tracking and updating of the interaction phase, thereby enhancing safety and comfort by reducing abrupt accelerations and decelerations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If the ego vehicle obtains a decision solution based on a predicted track of another vehicle and travels based on the decision solution, then the autonomous driving control can be implemented, but the interaction safety is poor and collision probability is high due to mismatch between prediction and actual behavior
Solution Approach 1:
The patent applies dynamics by making the decision solution adaptable and changeable during the interaction process. The system continuously updates the decision solution based on actual behavior of the another vehicle, transforming from a static prediction-based approach to a dynamic adaptation-based approach. This allows the ego vehicle to adjust its trajectory in real-time when the another vehicle's actual behavior deviates from the predicted track, thereby improving interaction safety while maintaining autonomous driving control.
2Device complexity
If the ego vehicle travels according to a previously obtained decision solution, then the autonomous driving process is simplified, but the ride experience deteriorates due to urgent braking when collision is detected
Solution Approach 1:
The patent implements feedback by continuously monitoring the actual behavior of the another vehicle and comparing it with the predicted track. When a deviation is detected, the system feeds this information back to update the decision solution, enabling smooth adaptive adjustments rather than abrupt emergency braking. This feedback mechanism maintains relatively simple decision-making logic while significantly improving ride experience by avoiding sudden stops.
3Ease of manufacture
If the ego vehicle uses a fixed prediction-based decision method, then the control policy is easy to implement, but the generalization capability is poor because different results are obtained in the same scenario
Solution Approach 1:
The patent applies parameter changes by dynamically modifying the decision solution parameters based on the actual behavior of the another vehicle. Instead of using a fixed prediction-based decision method, the system adjusts trajectory parameters in real-time according to observed deviations. This maintains ease of implementation while significantly improving generalization capability, as the control policy can adapt to various actual behaviors regardless of initial predictions.
Data Source
AI summary
This application discloses example intelligent driving methods and vehicles to which the example methods are applied. In one example method, a decision solution set of an ego vehicle may be obtained based on perception information, and a first decision solution used for a first interaction phase between the ego vehicle and a game target is obtained from the decision solution set. While controlling the ego vehicle to travel based on the first decision solution, a determination is made that the ego vehicle and the game target meet a condition for entering a second interaction phase. A second decision solution used for the second interaction phase between the ego vehicle and the game target is obtained from the decision solution set, and the ego vehicle is controlled to travel based on the second decision solution.