Bayesian Object Classifier for Autonomous Vehicle Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicles rely on human-generated rules for disambiguating object types classified by different sensor systems, which are prone to developer bias and errors due to subjective intuition, leading to inconsistent label assignments.
Innovation Solution
A Bayesian object classifier system that combines outputs from multiple object classifier modules, learned from labeled training data including sensor signals from various systems, to generate confidence score distributions and control vehicle operations, reducing bias and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If human-generated rules are used to disambiguate object types from different sensor systems, then the system can operate with simple logic, but the classification accuracy deteriorates due to developer bias and incorrect intuition
Solution Approach 1:
The patent replaces human-generated rule-based classification logic with a machine learning-based object classifier system. The system uses trained models that automatically learn optimal classification rules from data, substituting manual mechanical rule-creation with automated intelligent classification. This resolves the contradiction by eliminating developer bias while maintaining operational simplicity through automated decision-making.
Solution Approach 2:
The patent transforms the classification approach by changing from static human-defined rules to dynamic machine learning models that adapt parameters based on training data. The object classifier system adjusts classification thresholds and decision boundaries automatically through learning from labeled training data, improving accuracy without requiring complex manual rule configurations.
2Adaptability or versatility
If multiple object classifier modules are used to classify objects from different sensor signals, then the system can leverage diverse sensor data, but label consistency deteriorates when modules assign different labels to the same object
Solution Approach 1:
The patent merges outputs from multiple independent object classifier modules into a unified classification result. The system combines confidence scores and predictions from classifiers processing different sensor types (lidar, camera, radar) to produce a single consistent object label. This resolves the contradiction by maintaining diverse sensor input while achieving label consistency through integrated decision-making.
Solution Approach 2:
The patent implements feedback mechanisms where the system learns from discrepancies between different classifier modules. Through training on labeled data, the system adjusts individual module weights and thresholds to minimize conflicting labels. The feedback loop continuously optimizes the fusion strategy to maintain consistency while leveraging diverse sensor capabilities.
Data Source
AI summary
An autonomous vehicle is described herein. The autonomous vehicle includes several different types of sensor systems, such as image, lidar, and radar. The autonomous vehicle additionally includes a computing system that executes several different object classifier modules, wherein the object classifier modules are configured to identify types of objects that are in proximity to the autonomous vehicle based upon outputs of the sensor systems. The computing system additionally executes a Bayesian object classifier system that is configured to receive outputs of the object classifier modules and assign labels to objects captured in sensor signals based upon the outputs of the object classifier modules.