Robot Force Estimation Using Virtual Sensors and Regulatory Observers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining external interaction forces and moments in robots are inefficient, particularly in collision scenarios, as they rely on invasive sensors and are prone to errors due to dynamic coupling and singularity issues.
Innovation Solution
A procedure that combines real and virtual sensors using a regulatory observer to estimate external forces and moments, allowing for decoupled and configuration-independent measurements without the need for acceleration estimates, by modeling robot dynamics with passive mechanically bound joints and generalized coordinates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If real force and moment sensors are used to measure external interaction forces and moments, then measurement precision is improved, but device complexity increases and dynamic coupling errors occur
Solution Approach 1:
The patent combines real force/moment sensors with virtual sensors (dynamic model) into a unified measurement system. The virtual sensor is created through a dynamic model of the robot that computes expected forces and moments based on measured joint positions, velocities, and accelerations. By merging the real sensor data with the virtual sensor model, the system achieves accurate external force/moment measurement without requiring additional complex hardware, thus resolving the contradiction between measurement precision and device complexity
Solution Approach 2:
The patent introduces a regulatory observer as an intermediary computational system that processes sensor data and dynamic model information. This observer acts as a mediator between the physical sensors and the control system, computing external interaction forces and moments by comparing actual sensor readings with model predictions. The intermediary observer resolves measurement ambiguities and eliminates dynamic coupling errors without adding physical sensor complexity
2Loss of information
If acceleration measurements are used to compute external forces and moments, then measurement completeness is improved, but reliability decreases due to noise and computational errors
Solution Approach 1:
The patent implements a regulatory observer that uses feedback from real force/moment sensors to correct the virtual sensor model. The system continuously compares actual sensor measurements with model predictions and adjusts the estimation accordingly. This feedback mechanism allows the system to maintain reliable external force/moment estimation without depending solely on noisy acceleration measurements, thus resolving the contradiction between measurement completeness and reliability
Solution Approach 2:
The patent replaces direct reliance on acceleration measurements with a model-based virtual sensor approach. Instead of using potentially noisy acceleration data directly, the system substitutes a dynamic model that computes expected forces and moments from joint position, velocity, and torque data. This substitution reduces sensitivity to measurement noise while maintaining the necessary dynamic information for accurate external force/moment estimation
3Productivity
If dynamic model calculations are performed in real-time, then productivity is improved, but computational complexity increases causing singularity issues
Solution Approach 1:
The patent extracts and separates the computational tasks into distinct components: the virtual sensor model handles dynamic calculations, while the regulatory observer handles integration and correction. By taking out the computationally intensive dynamic model calculations and separating them from the control logic, the system achieves real-time performance without overwhelming computational complexity. This modular extraction allows efficient collision detection while avoiding singularity issues through proper task separation
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Method for determining external interaction forces and/or interaction moments of a robot (100) using at least one real force and/or moment sensor and a virtual model of the at least one force and/or moment sensor, wherein the at least one real force and/or moment sensor is modeled as a passive mechanically bound joint (112) together with robot dynamics and robot kinematics, and robot (100) and computer program product for executing such a method.