Google llc (20240248458). On-Robot Data Collection simplified abstract
Contents
On-Robot Data Collection
Organization Name
Inventor(s)
Ammar Husain of Palo Alto CA (US)
On-Robot Data Collection - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240248458 titled 'On-Robot Data Collection
The patent application describes systems and methods for improving the generation and selection of robot sensor data for manual annotation and training machine learning models used in robot operations.
- An on-robot controller can identify cross-modal inconsistencies, failed target tasks, or low confidence in model outputs to determine which sensor data should be sent for human annotation or model updating.
- Embedding vectors representing selected sensor data can be determined and transmitted to the robot for comparison with its own sensor data embeddings.
- If the embeddings are similar, the sensor data can be sent to the remote system for annotation or model updating.
Potential Applications: - Robotics industry for enhancing robot operations and efficiency. - Automation processes where machine learning models are used for decision-making.
Problems Solved: - Streamlining the process of selecting and sending sensor data for manual annotation or model updating. - Improving the accuracy and efficiency of machine learning models used in robot operations.
Benefits: - Enhanced performance and decision-making capabilities of robots. - Reduction in manual effort required for annotating sensor data. - Increased accuracy and reliability of machine learning models.
Commercial Applications: - Automation companies looking to improve their robot operations. - Robotics manufacturers seeking to enhance the capabilities of their robots.
Questions about the Technology: 1. How does the on-robot controller determine which sensor data should be sent for human annotation? 2. What are the potential implications of using embedding vectors for comparing sensor data in robots?
Original Abstract Submitted
systems and methods are provided for improved generation and selection of robot sensor data for manual annotation and/or use in training machine learning models used to operate robots. an on-robot controller can operate to determine a cross-modal inconsistency, that a temporally proximate target task was failed, and/or that a confidence in a model output indicate that particular sensor data should be transmitted to a remote system for human annotation and/or use in updating the machine learning model(s) of the robot. embedding vector(s) representing such selected sensor data (e.g., representing common aspects across a population of sets of sensor data) could also be determined and transmitted to the robot. the robot could then determine embeddings for sensor data and, if the embeddings are similar enough to the transmitted embedding(s), the sensor data could be transmitted to the remote system for annotation and/or model updating.
