Robotic Data Harvesting for Continuous Navigation Model Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robotic systems face challenges in achieving higher autonomy levels due to fears of malfunction or accidents, and existing data collection methods primarily focus on fixing flaws rather than enhancing navigation capabilities, limiting their ability to perform commercial tasks effectively.
Innovation Solution
A system for data collection and analysis that automatically gathers data from robotic operations, including malfunctions, and integrates it into a cloud server for trend evaluation and visualization, using convolutional neural networks to improve robotic navigation by identifying and addressing issues and enhancing perception models for semi-autonomous or fully autonomous navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data collection focuses on fixing flaws and preventing shortcomings, then system reliability improves, but navigation capability enhancement is limited
Solution Approach 1:
The system segments data collection into two distinct pipelines: one focused on flaw detection and prevention (reliability improvement) and another on general event collection for navigation enhancement. This segmentation allows both objectives to be pursued simultaneously without conflict, as each pipeline processes data according to its specific goals.
Solution Approach 2:
The data collection system is designed with multi-functionality to serve dual purposes: collecting data for flaw prevention while simultaneously gathering general operational events for navigation capability enhancement. The unified data collection infrastructure supports both reliability improvement and navigation enhancement through appropriate data selection and processing.
2Measurement precision
If manual data collection by human operators is used, then data accuracy improves, but productivity decreases
Solution Approach 1:
The system implements self-service through automated data collection mechanisms that operate without continuous human intervention. Sensors, logs, and system components automatically capture and transmit data to the cloud infrastructure, eliminating the need for manual data gathering while maintaining comprehensive coverage and accuracy through programmed collection criteria.
Solution Approach 2:
The system incorporates feedback mechanisms where collected data is automatically analyzed, processed, and used to improve future data collection strategies. The cloud-based infrastructure provides feedback loops that refine data selection criteria and enhance collection efficiency based on patterns identified in accumulated data, continuously improving both accuracy and productivity.
3Manufacturing precision
If comprehensive data is collected from all robotic operations, then model training quality improves, but data processing complexity increases
Solution Approach 1:
The system extracts only the necessary and relevant data from comprehensive robotic operations using predefined collection criteria and filters. Rather than processing all possible data, the system selectively extracts events and information that are most valuable for model training, reducing processing complexity while maintaining training quality through targeted data selection.
Solution Approach 2:
The cloud-based infrastructure serves as an intermediary between data collection and model training processes. This intermediary layer handles data aggregation, initial processing, filtering, and preparation, offloading complexity from the robotic systems themselves and enabling comprehensive data collection without proportionally increasing on-board processing requirements.
Data Source
AI summary
A system and method involves detecting a trigger event during operation of an autonomous ground vehicle traveling between two physical locations; generating event sequence data from primary sensor data, secondary sensor data, spatiotemporal data, and telemetry data through operation of a reporter; communicating the event sequence data to cloud storage and raw data to a streaming database; transforming the raw data into normalized data stored in a relational database through operation of a normalizer; operating a curation system to identify true trigger events from the normalized data and extract training data by way of a discriminator; operating a machine learning model within an active learning pipeline to generate a model update from aggregate training data generated from the training data by an aggregator; and reconfiguring the navigational control system with the model update communicated from the active learning pipeline to the autonomous ground vehicle.