Unified Track Confidence Model for Multi-Sensor Object Classification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing object detection and tracking systems in autonomous vehicles face challenges due to varying sensor data formats and differing detection algorithms across sensors, leading to inconsistent and potentially inaccurate object classifications and track confidences.

Innovation Solution

A combined track confidence and classification model that integrates data from multiple perception pipelines, including lidar, camera, and radar, to provide a unified track confidence metric and classification, reducing reliance on individual sensors and improving accuracy by aggregating and analyzing data from multiple sources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If multiple sensor types (lidar, camera, radar) with different detection algorithms are used, then the coverage and detection capability are improved, but the consistency and accuracy of object classifications and track confidences deteriorate due to varying sensor data formats and processing methods

Engineering Contradiction:
Improvedetection capabilityVSAvoidclassification consistency
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent combines track confidence metrics and classification results from multiple perception pipelines (lidar, camera, radar) into a unified model. This merging approach integrates diverse sensor data formats and processing methods while producing consistent classification outputs, resolving the contradiction between multi-sensor versatility and classification consistency.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The unified perception model serves multiple functions simultaneously: it processes different sensor types (lidar, camera, radar), generates track confidence metrics, produces object classifications, and ensures consistency across all sensor inputs. This multi-functional approach maintains detection capability while achieving classification consistency through a single integrated system.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If separate models are used for track confidence and classification, then the functionality is more comprehensive, but the system complexity and computational latency increase

Engineering Contradiction:
ImprovefunctionalityVSAvoidsystem complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges separate track confidence and classification models into a single unified perception model. This consolidation maintains comprehensive functionality by handling both track confidence assessment and object classification within one model, while reducing system complexity and computational overhead compared to running separate models.

Inventive Principle:
Principle #5Merging (Combining)

3Productivity

If individual sensor pipelines process data independently, then the processing is simpler and faster, but the reliability and robustness of object detection decrease due to inconsistent results

Engineering Contradiction:
Improveprocessing speedVSAvoiddetection reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The unified perception model acts as an intermediary that receives processed data from individual sensor pipelines and integrates their results. This intermediary approach preserves the processing efficiency of independent pipelines while ensuring detection reliability through unified confidence assessment and classification that reconciles inconsistent results from different sensors.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4107042B1Combined track confidence and classification model
Publication Date: 2026.04.01 ZOOX INC
  • EP4107042B1 patent drawingFigure 1
  • EP4107042B1 patent drawingFigure 2
  • EP4107042B1 patent drawingFigure 3

AI summary

Techniques are disclosed for a combined machine learned (ML) model that may generate a track confidence metric associated with a track and/or a classification of an object. Techniques may include obtaining a track. The track, which may include object detections from one or more sensor data types and/or pipelines, may be input into a machine-learning (ML) model. The model may output a track confidence metric and a classification. In some examples, if the track confidence metric does not satisfy a threshold, the ML model may cause the suppression of the output of the track to a planning component of an autonomous vehicle.