Autonomous Vehicle Object Tracking With Weighted Position Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in reliably detecting and tracking objects in their environment, particularly in complex urban settings, due to transient errors from sensor vibrations and temporary instability, which affect navigation accuracy.
Innovation Solution
An imagery-based tracking system that uses optical sensors to capture images of objects, combines detected positions with predicted positions based on historical data, and employs a weighted average to refine object location, integrating metadata from motion detection devices to stabilize tracking data and correlate new detections with historical tracks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor data is used to detect objects in the environment, then object detection capability is provided, but transient errors from sensor vibrations and temporary instability reduce tracking accuracy
Solution Approach 1:
The system performs preliminary actions by predicting object positions before actual detection occurs. The prediction module uses historical tracking data to forecast where objects will be at future time points, allowing the system to prepare for detection and reduce the impact of transient sensor errors by having expected positions to compare against actual measurements.
Solution Approach 2:
The system implements feedback mechanisms by continuously comparing detected object positions with predicted positions and using this information to refine future predictions. The weight assignment module uses the discrepancy between detected and predicted positions to dynamically adjust prediction weights, creating a closed-loop feedback system that improves tracking reliability over time despite transient sensor errors.
2Adaptability or versatility
If simple sensor detection is used, then device complexity is minimized, but the ability to continuously track and anticipate future actions of objects is insufficient
Solution Approach 1:
The tracking system is segmented into distinct functional modules: a prediction module that forecasts object positions, a detection module that captures actual sensor data, a weight assignment module that reconciles predictions with detections, and a tracking module that maintains object histories. This segmentation allows each module to perform a specific function, improving overall tracking capability while keeping individual module complexity manageable.
Solution Approach 2:
The prediction module serves multiple functions: it anticipates future object positions for proactive tracking, provides reference points for detecting transient errors, and generates predictions that can be weighted and combined with actual detections. This multi-functionality increases system adaptability without requiring separate dedicated components for each function.
3Measurement precision
If detected position is used alone, then tracking responds to current state, but transient errors reduce accuracy; if predicted position is used alone, then future positioning is anticipated, but deviation from actual position occurs
Solution Approach 1:
The system performs preliminary position predictions before actual detection occurs, allowing it to anticipate where objects will be rather than merely reacting to where they are. This preliminary action enables the system to maintain accurate tracking even during transient sensor errors, as the predicted position serves as a reference that can compensate for temporary detection failures.
Solution Approach 2:
The system dynamically changes parameters by adjusting the weight assigned to predicted versus detected positions based on their relative reliability. When detection quality is high, the weight on detected position increases; when transient errors are detected or prediction confidence is high, the weight shifts toward predicted position. This dynamic parameter adjustment optimizes position accuracy without sacrificing response time.
Data Source
AI summary
This disclosure relates in general to systems and methods for optically tracking objects proximate an autonomous vehicle. In particular, an object tracking system capable of refining position data for objects being tracked by determining a location of the objects surrounding the autonomous vehicle at least on part on previously determined locations of the objects. In certain instances, the predicted and detected locations used to arrive at a refined location for the objects can be weighted in different ways depending on conditions of the sensor data and quality of the historical data.