Stereoscopic Camera Robot Force Torque Sensor Non-Linearity Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current surgical stereoscopic cameras connected to robotic arms face challenges due to the non-linearity characteristics of force/torque sensors, which affect the control of robotic arms, especially in sensitive medical applications.
Innovation Solution
A system comprising a robotic arm with a 6DOF force/torque sensor, a processor, and memory, which generates real-time sensor corrections for offset, linear, and non-linear deviations across multiple axes, allowing for accurate calibration and correction of sensor non-linearity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a force/torque sensor is used to sense operator manipulation of the stereoscopic camera, then the robotic arm can be controlled collaboratively with effortlessness, but the sensor exhibits non-linearity characteristics that affect control accuracy
Solution Approach 1:
The system performs preliminary calibration by collecting sensor data across multiple orientations and positions of the robotic arm before actual operation. A correction model is trained in advance using this calibration data to compensate for non-linearity effects, enabling accurate control without real-time computational overhead during collaborative manipulation.
Solution Approach 2:
The system changes the parameters used for sensor interpretation by applying correction factors derived from calibration data. Instead of using raw sensor readings directly, the system transforms the sensor data through a learned correction model that accounts for non-linearity, effectively changing the measurement parameters to achieve both ease of operation and measurement precision.
2Measurement precision
If re-zeroing of the sensor is performed to compensate for non-linearity, then some accuracy can be restored, but it cannot occur in real-time and is ineffective when force translates between axes
Solution Approach 1:
The system performs comprehensive calibration in advance by collecting sensor data across the full range of motions and orientations. This preliminary action creates a correction model that can be applied instantly during operation, eliminating the need for slow re-zeroing procedures while maintaining accuracy even when forces translate between axes.
Solution Approach 2:
The system introduces an intermediary correction model that sits between the raw sensor and the control system. This intermediary layer processes sensor readings through a pre-trained model that accounts for non-linearity and cross-axis effects, providing accurate real-time corrections without requiring direct intervention or re-zeroing of the sensor during operation.
3Measurement precision
If neural networks are used to correct sensor non-linearity, then effective correction can be achieved, but extensive computational resources and data requirements make it impractical
Solution Approach 1:
The system applies partial action by using a simplified correction approach that captures the essential non-linearity effects without requiring a full neural network. The correction model uses a subset of the computational complexity of a complete neural network while still achieving effective correction for the specific application domain of robotic arm sensor calibration.
Solution Approach 2:
The system changes the parameters of the correction approach by using a lightweight model with fewer parameters than a full neural network. The correction model uses optimized parameters derived from calibration data that provide sufficient accuracy for sensor non-linearity compensation without requiring extensive computational resources or large datasets for training.
Data Source
AI summary
New and innovative systems and methods for calibrating and correcting sensors associated with a collaborative robot are disclosed. An example system comprises: at least one robotic arm; a sensor affixed to a location on the robotic arm, wherein the sensor measures force and torque across six degrees of freedom (6DOF); a processor; and memory. The system may receive, from the sensor, sensor input in real-time that indicate a measured force or torque. The system may generate, in real-time, sensor corrections that correspond to offset, linear, and non-linear deviations of the measured force in each sensor axis. The sensor corrections may correspond to offset, linear, and non-linear cross-coupling of the measured force between two or more sensor axes. The sensor corrections may be determined by applying offset, linear, and non nonlinear corrections to each degree of freedom (DOF) from every other DOF.