EPD Confidence Propagation Across ADAS Functional Modules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automotive applications, such as ADAS and DMS systems, face challenges in providing reliable outputs due to noisy measurements and the lack of confidence measures in deterministic outputs, leading to uncertain estimates and potential safety issues when making decisions based on these outputs.
Innovation Solution
An entropy of predictive distribution (EPD)-based confidence system is implemented, using multiple functional modules with associated EPD modules to generate and combine confidence measures, allowing for unified representations of confidence across different types of outputs, even with high-dimensional input data, and providing a single, efficient confidence value independent of scale.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deterministic outputs are used without confidence measures, then the system operation is simple, but the reliability of decisions is reduced
Solution Approach 1:
The system segments the output generation into multiple functional modules (e.g., object detection module, trajectory prediction module, risk assessment module), where each module independently generates confidence measures for its specific output. This allows confidence assessment to be added without requiring complete system redesign, resolving the contradiction between reliability improvement and system complexity.
Solution Approach 2:
An entropy of predictive distribution (EPD) module acts as an intermediary between functional modules and decision-making processes. The EPD module computes confidence measures from predictive distributions and combines them through logical operations, serving as a mediator that translates complex probabilistic outputs into actionable confidence scores without adding excessive complexity to the overall system.
2Reliability
If confidence measures are generated for each functional module, then the reliability of outputs is improved, but the device complexity increases
Solution Approach 1:
Multiple confidence measures from different functional modules are merged into a unified confidence score through logical combinations (e.g., combining object detection confidence with trajectory prediction confidence). This merging approach maintains comprehensive reliability assessment while reducing the number of separate confidence outputs the system must handle, balancing reliability improvement with complexity management.
Solution Approach 2:
The EPD-based confidence measure framework is designed as a universal mechanism that can be applied across different functional modules (object detection, trajectory prediction, risk assessment). This multi-functional approach allows a single confidence generation methodology to serve multiple purposes, reducing the need for module-specific confidence mechanisms and thereby limiting complexity growth.
3Adaptability or versatility
If heterogeneous functional modules are used to generate different types of output data, then the adaptability of the system is improved, but the difficulty of generating unified confidence measures increases
Solution Approach 1:
The system transforms diverse confidence measures from heterogeneous functional modules into a unified parameter space using entropy of predictive distribution. By converting different types of confidence scores (object detection confidence, trajectory prediction confidence) into EPD values, the system maintains adaptability to handle different output types while simplifying the unification process through parameter transformation.
Data Source
AI summary
A method includes performing data processing operations using multiple functional modules. Each functional module is configured to perform one or more data processing operations in order to process input data and generate output data. The method also includes, for each functional module, generating a confidence measure associated with the output data generated by the functional module. At least two of the functional modules are configured to operate logically sequentially such that (i) a first of the functional modules provides the output data generated by the first functional module to a second of the functional modules and (ii) the confidence measure associated with the output data generated by the second functional module is based at least partially on the confidence measure associated with the output data generated by the first functional module. The multiple functional modules include heterogeneous functional modules configured to generate different types of output data.