Robot Actuator Weight Inference via AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robots without scale or weight sensors face challenges in determining the weight or quantity of retrieved objects, as they lack direct measurement capabilities.
Innovation Solution
The use of artificial intelligence and machine learning techniques to derive object weight or quantity from measurements obtained from robot actuators and associated sensors, such as vacuum pressure, carriage pull force, lift displacement, and retrieval element displacement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If robots are equipped with scale or weight sensors to directly measure object weight, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent replaces direct mechanical weight sensing with an AI/ML-based inference system that processes measurements from existing actuator systems. The neural network model analyzes actuator activation patterns, forces, and movements to derive weight information without requiring dedicated weight sensors, thus substituting a mechanical sensing approach with an intelligent computation approach.
Solution Approach 2:
The patent makes existing actuator systems serve dual purposes: their primary function for object manipulation and retrieval, and a secondary function for weight measurement. By training the AI model to interpret actuator behavior, the same hardware components perform multiple functions, eliminating the need for separate weight sensing equipment.
2Device complexity
If robots use existing actuators for weight measurement, then device complexity is reduced, but measurement precision deteriorates due to indirect measurement
Solution Approach 1:
The patent introduces an AI/ML inference system as an intermediary between the actuator measurements and the weight determination. The neural network model acts as a mediator that processes complex, indirect actuator data and transforms it into accurate weight estimates, bridging the gap between indirect measurements and precise weight information.
Solution Approach 2:
The patent transforms the measurement parameters by training the AI model to recognize patterns in actuator activation data that correlate with weight. The system changes from directly measuring weight to measuring actuator responses and inferring weight through learned parameter relationships, enabling accurate measurement through parameter transformation.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables robots to accurately determine the weight or quantity of objects without the need for weight sensors, by converting actuator activation measurements into reliable weight or quantity metrics.
Implementation Method 1
a vacuum to engage, retrieve, place, or otherwise interact with the object
Implementation Method 2
a lift to raise and lower a platform
Implementation Method 3
drive motors to move the robot across a ground surface
Data Source
AI summary
Disclosed are systems and methods for computing an object's weight or the quantity of items within the object based on activations of different robot actuators used in retrieving the object from a storage location. Specifically, the robot may activate one or more actuators during retrieval of an object, may obtain a measurement in response to activating a particular actuator, and may derive a weight of the object based on the measurement by converting the measurement from a first range of values that are associated with activations of the particular actuator to a second range of values that are disassociated with activations of the particular actuator. The robot may then modify its operation in response to the weight that is derived for the object matching or being mismatched to an expected or last tracked weight.