Sensor-Based Reference Map Updates Using Self-Supervised Change Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automating the update of sensor-based maps for autonomous driving systems is challenging due to difficulties in quickly and accurately identifying change detections, which can lead to performance slowdowns and safety hazards.
Innovation Solution
A method using a machine-learned model trained with self-supervised learning to determine differences between sensor data and reference map features, enabling quick and accurate identification of change detections and triggering map updates, allowing vehicles to operate safely in autonomous mode.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If automation attempts are made to identify all possible change detections, then map completeness is improved, but processing time and computational resources increase significantly
Solution Approach 1:
The system uses self-supervised learning where the model learns to identify change detections autonomously from sensor data without requiring extensive manual labeling or supervision. The model serves itself by learning from the data patterns it encounters, enabling automated change detection while reducing the need for resource-intensive manual verification processes
Solution Approach 2:
The patent changes the operational parameters of the learning system by using self-supervised learning instead of traditional supervised learning approaches. This parameter change in the learning methodology allows the system to process data more efficiently while maintaining high accuracy in change detection, resolving the contradiction between completeness and processing time
2Measurement precision
If traditional supervised learning methods are used for change detection, then accuracy can be maintained, but processing speed and resource efficiency deteriorate
Solution Approach 1:
The patent replaces traditional supervised learning mechanisms with self-supervised learning mechanisms. This substitution changes the fundamental approach from requiring labeled training data and extensive computational processing to autonomously learning from unlabeled sensor data, thereby improving processing speed while maintaining accuracy in change detection
3Reliability
If manual verification processes are used for change detections, then map accuracy is improved, but system complexity and operational overhead increase
Solution Approach 1:
The system eliminates the need for manual verification by implementing self-supervised learning that autonomously identifies and verifies change detections. The model learns to distinguish between actual environmental changes and sensor noise or errors through self-supervision, thereby maintaining map accuracy while reducing system complexity and operational overhead associated with manual verification processes
Data Source
AI summary
This document describes change detection criteria for updating sensor-based maps. Based on an indication that a registered object is detected near a vehicle, a processor determines differences between features of the registered object and features of a sensor-based reference map. A machine-learned model is trained using self-supervised learning to identify change detections from inputs. This model is executed to determine whether the differences satisfy change detection criteria for updating the sensor-based reference map. If the change detection criteria is satisfied, the processor causes the sensor-based reference map to be updated to reduce the differences, which enables the vehicle to safely operate in an autonomous mode using the updated reference map for navigating the vehicle in proximity to the coordinate location of the registered object. The map can be updated contemporaneously as changes occur in the environment and without hindering performance, thereby enabling real-time awareness to support controls and to improve driving-safety.