Obstacle Intent Prediction at Intersections for Faster Autonomous Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face challenges in accurately predicting the intent of moving obstacles, leading to delays in decision-making and increased risk of collisions due to delayed response and imperfect perception of motion status and future behavior.
Innovation Solution
Determine the intent of a target obstacle by analyzing its historical motion status to calculate probability distributions for cutting across traffic or yielding at an intersection, using historical data and current motion status to make accurate predictions and control the vehicle's actions accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the autonomous driving system uses traditional detection and prediction methods for moving obstacles, then the system structure is simple, but the response delay increases and collision risk increases
Solution Approach 1:
The patent applies preliminary action by predicting the intent of moving obstacles (cutting across traffic or yielding) before the actual interaction occurs. The system analyzes historical motion status and calculates probability distributions of future behaviors in advance, enabling the autonomous vehicle to prepare decision-making and planning before the obstacle reaches critical proximity, thus reducing response delay and improving safety
Solution Approach 2:
The patent implements beforehand cushioning by introducing a buffer time through intent prediction. By determining the probability distributions of obstacle behaviors in advance and comparing them with current motion status, the system creates a temporal cushion that allows for earlier decision-making, compensating for the inherent response delays in vehicle chassis and control systems
2Measurement precision
If the autonomous driving system performs detailed detection and prediction of moving obstacle behavior, then the measurement precision improves, but the computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the complex prediction problem into distinct components: historical motion status analysis, probability distribution calculation for cutting across traffic, probability distribution calculation for yielding, and current status comparison. This modular approach improves measurement precision while managing computational complexity through structured processing
Solution Approach 2:
The patent uses parameter changes by transforming raw motion status data into probability distributions that represent different behavioral intents. By changing the parameter representation from simple position-speed data to probabilistic intent models, the system achieves higher detection accuracy in predicting obstacle behavior while maintaining computational feasibility through mathematical transformations
3Measurement precision
If the autonomous driving system makes decisions based on current motion status only, then the decision-making speed is fast, but the prediction accuracy of obstacle intent is insufficient
Solution Approach 1:
The patent applies preliminary action by pre-calculating probability distributions of obstacle behaviors based on historical motion status before the actual decision moment. This allows the system to have prediction results ready in advance, achieving high intent prediction accuracy without sacrificing decision-making speed when the current status is compared against pre-computed distributions
Solution Approach 2:
The patent implements continuity of useful action by continuously updating and maintaining the probability distributions as the obstacle moves. The system continuously processes historical motion status data and maintains ready-to-use prediction models, ensuring both high prediction accuracy and rapid decision-making through uninterrupted computational flow
Data Source
AI summary
Methods and devices are provided for determining an intent of a target, applicable to intelligent driving. An example method includes: determining, based on a historical motion status of a target obstacle, a probability distribution of a motion status in which the target obstacle cuts across traffic to pass through an intersection point and a probability distribution of a motion status in which the target obstacle yields to pass through the intersection point; and determining in advance, based on a current motion status of the target obstacle and these probability distributions, whether the target obstacle cuts across traffic or yields to pass through the intersection point. Embodiments can be applied to intelligent vehicles (e.g., autonomous driving). Before colliding with the target obstacle, an intent of the obstacle is identified or calculated, enabling route planning for safety.