Wearable Sensor Glove for Scalable Robotic Hand Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for collecting training data for robotic hand manipulators are inefficient, lack scalability, and require dedicated training to control the robot, leading to poor usability and resource wastage.
Innovation Solution
A sensorised device configured to be worn by a human collects training data without the need for the robotic hand manipulator to be present, using a sensor system identical to that of the manipulator, allowing for efficient data collection and interoperability with the manipulator.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a controller is attached to the robotic hand manipulator and carried by a human for data collection, then training data can be collected for task automation, but the system requires dedicated training to control the robot, leading to poor usability and loss of time
Solution Approach 1:
The patent creates a digital twin of the robotic hand manipulator that runs simulations instead of requiring physical robot presence. This copying approach allows data collection without the complexity of controlling the actual robot, eliminating the need for dedicated training while maintaining data quality through realistic simulation environments.
Solution Approach 2:
The patent replaces the mechanical control system (physical robot manipulation) with a software-based simulation system. Instead of requiring humans to physically control the robotic hand through controllers, the system uses virtual simulations that automatically generate training data, substituting mechanical interaction with computational modeling.
2Reliability
If a controller is attached to the robotic hand manipulator for data collection, then training data can be obtained, but the approach lacks scalability and efficiency due to resource usage requirements
Solution Approach 1:
By creating virtual copies of the robotic system through digital twins, the patent enables parallel simulation of multiple scenarios simultaneously without requiring additional physical robots. This multiplying effect through copying allows extensive data collection across diverse conditions while using minimal physical resources, dramatically improving productivity and scalability.
Solution Approach 2:
The patent performs preliminary actions by pre-simulating various task scenarios and conditions in the virtual environment before physical deployment. This advance preparation generates comprehensive training data sets that reduce the need for extensive physical testing, improving efficiency by doing the heavy lifting in the computational domain first.
3Reliability
If a controller is attached to the robotic hand manipulator, then training data can be collected, but the device complexity increases and requires dedicated hardware
Solution Approach 1:
The patent replaces complex physical hardware (robotic hand manipulator with attached controllers) with a software-based digital twin. This virtual copy replicates the functional behavior of the physical system without requiring the actual hardware to be present during data collection, dramatically reducing device complexity while maintaining data collection capability.
Solution Approach 2:
The patent extracts the essential data collection function from the physical robot system and isolates it into a separate simulation environment. By taking out the core functionality (data generation) from the complex hardware platform, the system can collect training data without requiring the robotic manipulator or its controllers to be physically present, simplifying the overall system architecture.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed is a method for collecting training data for a robotic hand manipulator for automating a manual task of a human. Furthermore, a method for automating a manual task is disclosed. Finally, corresponding devices and systems are disclosed.