Object Validation System for ADS Threat Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems (ADS) face challenges in accurately validating perceived surrounding objects for safety-critical threat assessment, leading to potential false positives and negatives during emergency maneuvers, which can result in unnecessary braking or collisions.
Innovation Solution
An object validation system that stores sensor data from various modalities in separate buffers, determines object data using a perception module, and evaluates this data against predeterminable matching criteria to classify objects as validated or unvalidated, ensuring accurate threat assessment for emergency maneuvers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If perception algorithms are made more sensitive to detect more objects, then detection coverage is improved, but false positive rate increases leading to unnecessary emergency braking
Solution Approach 1:
The patent introduces an object validation system as an intermediary layer between the perception module and emergency maneuver logic. This validation system uses multiple validation techniques (consistency validation, persistence validation, plausibility validation) to verify perceived objects before triggering emergency maneuvers, thereby reducing false positives while maintaining detection coverage
Solution Approach 2:
The system implements feedback mechanisms where validation results are fed back to adjust perception thresholds and validation parameters. The emergency maneuver logic provides feedback about actual outcomes (true positives vs false positives) to continuously optimize the balance between detection coverage and false positive rate
2Object-generated harmful factors
If perception algorithms are made more conservative to reduce false positives, then false positive rate decreases, but detection coverage reduces leading to missed threats
Solution Approach 1:
The validation parameters and thresholds are made dynamic rather than static. The system adjusts validation stringency based on contextual factors such as object type, location, motion characteristics, and environmental conditions, allowing high detection coverage for critical threats while maintaining low false positive rates for ambiguous objects
Solution Approach 2:
The patent segments the validation process into multiple independent validation stages (consistency check, persistence check, plausibility check). Each stage can be adjusted independently, allowing the system to apply stricter validation to high-risk object categories while using more lenient validation for low-risk categories, thereby maintaining both high detection coverage and low false positive rate
3Measurement precision
If multiple validation techniques are applied to verify perceived objects, then accuracy is improved, but computational complexity increases
Solution Approach 1:
The system applies validation techniques selectively rather than uniformly to all perceived objects. Low-risk objects undergo minimal validation, while high-risk objects undergo full multi-stage validation. This partial action approach maintains high accuracy for critical cases while reducing overall computational complexity
Solution Approach 2:
The patent implements preliminary filtering stages that quickly eliminate obviously invalid objects before applying more computationally intensive validation techniques. Objects that fail simple consistency checks are rejected early, preventing unnecessary computational expenditure on clearly false positives
4Object-generated harmful factors
If validation thresholds are set high to ensure safety, then false positives are reduced, but response time increases due to additional validation steps
Solution Approach 1:
Validation thresholds and required validation depth are dynamically adjusted based on threat level assessment. For objects exhibiting high-risk characteristics (sudden appearance, abnormal motion, critical location), the system reduces validation requirements and lowers thresholds to enable faster response. For low-risk objects, higher thresholds and more validation steps are applied to reduce false positives
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure relates to a method performed by an object validation system (1) for validating perceived surrounding objects to support safety-critical threat assessment governing emergency maneuvering of an ADS (21) on-board a vehicle (2). The object validation system stores (1001) in respective sensor/modality-specific data buffers, respective sensor/modality-specific sensor data (3) obtained at least during a predeterminable time interval continuously and/or intermittently from one or more vehicle-mounted surrounding detecting sensors (23). The object validation system further determines (1002) with support from a perception module (22) configured to generate perception data (220) based on sensor data from one or more vehicle-mounted surrounding detecting sensors, object data (40) of a perceived object (4) valid for said time interval. Moreover, the object validation system evaluates (1003) one or more of the respective sensor/modality-specific data buffers, separately, in view of the object data. Furthermore, the object validation system determines (1004) that the perceived object is a validated object when the object data matches sensed objects in the one or more respective sensor/modality-specific data buffers according to predeterminable matching criteria, and otherwise is an unvalidated object. The disclosure also relates to an object validation system in accordance with the foregoing, a vehicle comprising such an object validation system, and a respective corresponding computer program product and non-volatile computer readable storage medium.