Vehicle Warning Logic for Hidden Road User Collision Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing collision warning systems for vehicles are limited in their ability to accurately assess collision risks in diverse road situations, leading to irrelevant alerts that users may ignore, and fail to address a wide range of potential hazards.
Innovation Solution
A warning system for vehicles that detects pairs of road users who cannot see each other, predicts their trajectories based on lane following and traffic rules, calculates collision probabilities, and triggers alerts only when certain conditions are met to ensure relevance and effectiveness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a simple risk evaluation system is used to detect potential collisions, then the system complexity is reduced and ease of operation is improved, but the measurement precision of collision risk assessment deteriorates leading to false alerts
Solution Approach 1:
The risk evaluation is segmented into multiple independent modules: detection device for identifying road users, trajectory prediction device for forecasting paths, collision risk assessment device for calculating probabilities, and alert control device for determining alert necessity. This segmentation allows each module to specialize in one function, improving overall measurement precision while maintaining operational simplicity through modular architecture.
Solution Approach 2:
The system performs preliminary trajectory prediction for multiple road users before collision detection. By pre-calculating predicted trajectories and potential intersection points, the system prepares collision risk assessments in advance, enabling more accurate real-time evaluation without increasing operational complexity during critical moments.
2Measurement precision
If trajectory prediction taking lanes is used for moving objects following lanes, then the measurement precision of collision risk assessment is improved, but the device complexity increases
Solution Approach 1:
The system applies different levels of trajectory prediction complexity based on local conditions: for moving objects following lanes, it uses lane-constrained trajectory prediction with higher precision; for other objects, it uses simpler motion models. This local quality approach optimizes measurement precision where needed while controlling overall device complexity through conditional processing.
Solution Approach 2:
The trajectory prediction device dynamically adjusts its modeling approach based on object characteristics and environmental context. When objects follow lanes, the system activates lane-based prediction algorithms; otherwise, it uses general motion models. This dynamic adaptation allows the system to maintain high measurement precision for relevant cases while managing device complexity through selective algorithm activation.
3Reliability
If the system warns in all detected risky situations, then the reliability of safety warning is improved, but the loss of information increases due to irrelevant alerts
Solution Approach 1:
The alert control device receives feedback from multiple sources: detected risky situations, predicted trajectories, collision probabilities, and host vehicle position. By integrating this feedback loop, the system selectively generates alerts only when comprehensive analysis confirms both collision risk and potential driver benefit, improving reliability while filtering out irrelevant warnings that would cause information loss.
Solution Approach 2:
Before issuing alerts, the system performs preliminary evaluation of alert relevance by assessing host vehicle position relative to the risky situation and predicting whether the driver can benefit from the warning. This preliminary action filters out situations where alerts would be irrelevant, maintaining high reliability for genuine risks while preventing information loss from unnecessary warnings.
Data Source
Figure 1~2
Figure 3~4
Figure 5A
AI summary
When two target objects cannot see each other and at least one of them is a moving object following lanes, the warning system identifies the couples of two predicted trajectories having at least one point in common or an inter-trajectory distance less than a distance threshold (dth), among all possible couples of two predicted trajectories for the two respective target objects, then calculates a probability of collision for each identified couple of trajectories, by determining a predicted collision time and calculating probability distribution functions for the positions of the two target objects at said predicted collision time, and checks whether or not the probability of collision calculated for at least one of the identified couples of trajectories is more than a collision probability threshold in order to trigger an alert towards at least one of the two target objects.