Emergency Collision Avoidance Strategy Using Predicted Vehicle Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle safety technologies lack a method for continuously avoiding risky situations during autonomous driving, particularly in emergency collisions, beyond minimal risk maneuvers.
Innovation Solution
A system and method utilizing a deep learning-based approach to predict vehicle trajectories, determine collision risk, and generate a bird's eye view image to select an avoidance strategy, including lane changes, steering to shoulder stops, and evasive steering, using convolutional neural networks to output the optimal strategy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deep learning-based trajectory prediction and avoidance strategy generation are implemented, then collision avoidance capability is improved, but system complexity increases
Solution Approach 1:
The system segments the collision avoidance function into multiple independent modules: risk degree determination unit, vehicle trajectory prediction unit, avoidance trajectory prediction unit, and collision avoidance strategy determination unit. Each module handles a specific aspect of the avoidance process, making the complex system manageable and maintainable while achieving high reliability through specialized processing in each segment.
Solution Approach 2:
The system performs preliminary actions by predicting future vehicle trajectories and avoidance trajectories before actual collision occurs. The risk degree determination unit continuously assesses potential collisions in advance, and multiple avoidance strategies are pre-calculated and evaluated, allowing the system to select the optimal strategy before emergency situations arise.
2Reliability
If multiple avoidance strategies are predicted and evaluated, then collision avoidance effectiveness is improved, but computation time increases
Solution Approach 1:
The system applies partial action by generating multiple avoidance strategies (excessive) but only selecting and executing the single best strategy (partial). The avoidance trajectory prediction unit generates several candidate trajectories, and the collision avoidance strategy determination unit evaluates them to choose the optimal one, avoiding the need to execute all predicted strategies while ensuring the best possible outcome.
Solution Approach 2:
Multiple avoidance strategies are pre-calculated and stored before emergency situations occur. The system maintains a library of potential avoidance maneuvers with their associated cost functions pre-computed, allowing for rapid selection during critical moments without performing full real-time calculations for each possible strategy.
3Loss of information
If bird's eye view generation with trajectory predictions is implemented, then situation awareness is improved, but processing load increases
Solution Approach 1:
The bird's eye view generation is segmented into distinct layers: current state representation, predicted vehicle trajectory overlay, and predicted avoidance trajectory overlay. Each layer processes and displays specific information separately, allowing the system to manage complex visual information efficiently while providing comprehensive situation awareness through the composite view.
Data Source
AI summary
A system for establishing a strategy for allowing a vehicle to avoid collision risk. The system includes a risk degree determination unit configured to determine a degree of collision risk with a nearby object of an ego vehicle; a vehicle trajectory prediction unit configured to predict a future vehicle trajectory of the nearby object; an avoidance trajectory prediction unit configured to predict a future avoidance trajectory of the ego vehicle corresponding to a plurality of avoidance strategies for avoiding collision with the nearby object; and a collision avoidance strategy determination unit configured to learn the future vehicle trajectory of the nearby object and the future avoidance trajectory of the ego vehicle and output an avoidance strategy selected from the plurality of avoidance strategies.