Autonomous Vehicle Obstacle Avoidance Using Reliability Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in safely navigating unexpected obstacles that have not been pre-classified, leading to unnecessary vehicle control and potential safety risks due to erroneous sensor recognition.
Innovation Solution
An apparatus and method that utilize sensors like cameras, lidars, and radars to determine the reliability of an unclassified object, with a processor controlling the vehicle to avoid the object only if the reliability value exceeds a threshold, thereby reducing unnecessary vehicle control and ensuring safety.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If autonomous vehicles avoid all unclassified objects, then safety is improved, but unnecessary vehicle control and traffic flow disruption occur
Solution Approach 1:
The patent changes the parameter of object classification from binary (classified/unclassified) to a reliability score continuum. By calculating reliability values based on multiple sensor data characteristics and comparing them against threshold values, the system distinguishes between truly dangerous unclassified objects and harmless ones, enabling selective avoidance that maintains safety while reducing unnecessary maneuvers and preserving traffic flow.
2Reliability
If autonomous vehicles avoid all unclassified objects, then safety is improved, but ride quality deteriorates due to unnecessary evasive maneuvers
Solution Approach 1:
The system transforms the safety decision parameter from a binary classification to a graded reliability assessment. By evaluating multiple sensor characteristics and computing reliability scores, the vehicle can distinguish between objects requiring avoidance and those that do not, thereby eliminating unnecessary evasive maneuvers that would degrade ride quality while maintaining adequate safety through threshold-based decision making.
3Quantity of substance
If sensors recognize all objects, then detection completeness is improved, but measurement precision deteriorates due to erroneous recognition
Solution Approach 1:
The patent implements a feedback mechanism where reliability scores are calculated based on sensor data characteristics and used to adjust the avoidance decision. The system continuously evaluates the reliability of unclassified object detections and compares them against threshold values, providing feedback that prevents erroneous recognition from triggering unnecessary avoidance actions while maintaining complete detection of all objects.
Solution Approach 2:
The system changes the measurement approach from binary object identification to continuous reliability scoring. By evaluating multiple sensor characteristics and computing reliability values, the system maintains detection completeness for all objects while improving identification accuracy through graded assessment and threshold-based filtering of false positives.
Data Source
AI summary
The present disclosure relates to a device and a method for avoiding an obstacle. The device includes a sensor that detects an object outside a vehicle, a processor, and memory storing instructions. The device determines whether the object belongs to any pre-classified type, determines a reliability value associated with information on the object based on the object not belonging to any pre-classified type, and controls, based on the reliability value being greater than a threshold reliability value, the driving controller to avoid the object.