Surrounding Object Path Prediction for Abnormal Obstacle Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Advanced Driver Assistance Systems (ADAS) struggle to recognize and avoid abnormal dynamic obstacles, and accurately predict the future movement path of dynamic objects to determine collision possibilities with vehicles.
Innovation Solution
A method and system that utilize sensors to track the movement path of objects, generate prediction path data using a learned artificial neural network model, and adjust the vehicle's driving trajectory to prevent collisions, incorporating road maps and probability calculations to determine the likelihood of object movement paths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If ADAS uses traditional obstacle detection methods, then it can avoid static obstacles, but it fails to recognize and avoid abnormal dynamic obstacles
Solution Approach 1:
The system transitions from static obstacle detection to dynamic path prediction by continuously tracking moving objects and forecasting their future positions. The prediction module generates multiple possible paths for dynamic obstacles, enabling the system to adapt to abnormal movement patterns that traditional methods cannot anticipate.
Solution Approach 2:
The system performs preliminary path prediction for dynamic obstacles before actual collision risk materializes. By forecasting potential movement paths and calculating collision risks in advance, the system can prepare avoidance maneuvers earlier, improving both reliability and adaptability to dynamic situations.
2Measurement precision
If ADAS tracks dynamic obstacles, then it can identify moving objects, but it cannot accurately predict their future movement paths
Solution Approach 1:
The prediction module acts as an intermediary between current tracking data and future collision risk assessment. It processes observed movement patterns and generates probable future paths, bridging the gap between present tracking capability and future collision prediction needs.
Solution Approach 2:
The system changes the parameter representation from current position-only tracking to include multiple predicted future position parameters. By generating multiple possible paths with associated probabilities, the system preserves information about uncertain future movements that single-point prediction would lose.
3Measurement precision
If the system generates multiple prediction paths, then it improves collision prediction accuracy, but it increases computational complexity
Solution Approach 1:
The system generates multiple prediction paths (excessive action) to ensure comprehensive coverage of possible obstacle movements. This approach prioritizes prediction accuracy over computational efficiency, creating more paths than strictly necessary to capture the full range of potential obstacle behaviors.
Solution Approach 2:
The prediction system segments the future movement space into multiple discrete paths, each with its own probability. This segmentation allows the collision risk calculation to process complex uncertainty by breaking it down into manageable path segments rather than treating it as a continuous problem.
Data Source
AI summary
A method predicting a path of an object includes: recognizing the object by using at least one sensor of the vehicle; generating movement path data associated with the object by tracking a movement path of the object during a first time interval when the object is recognized; and generating prediction path data including at least one prediction path associated with the object during a second time interval based on the generated movement path data.