Whisker Sensor 3D Mapping via Neural Network

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current robotic systems for three-dimensional (3-D) feature extraction and tactile sensing, such as those using whisker arrays, face limitations in accurately determining object shape, texture, and material compliance due to complexities in interpreting sensory data from diverse surface features and varying friction coefficients.

Innovation Solution

The use of neural networks, trained via evolutionary algorithms, to process moment and force data from tactile sensors, allowing for the reconstruction of 3-D morphology, position, orientation, and compliance of objects by learning from training data and adapting to new surfaces, incorporating techniques like contact point estimation and compliance mapping.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional whisker arrays are used for 3-D feature extraction, then the system structure remains simple, but the accuracy of determining object shape and texture deteriorates due to inability to handle diverse surface features and varying friction coefficients

Engineering Contradiction:

Improveaccuracy of determining object shape and textureVSAvoidcomplexity of interpreting sensory data

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

A neural network is introduced as an intermediary between the whisker sensors and the object characterization system. The neural network processes complex sensory data from the whisker array, mapping moment and force measurements to accurate 3-D surface topography and material compliance representations, thereby resolving the contradiction between simple sensor structure and accurate measurement

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system transitions from direct mechanical measurement to a parameter transformation approach using neural networks. By training the network on diverse surface features and friction conditions, the system learns to represent objects in terms of their topographic and compliance parameters, improving measurement accuracy without increasing physical sensor complexity

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If neural networks with evolutionary algorithms are used to process sensory data, then the accuracy of 3-D surface mapping improves, but the computational complexity and training time increase

Engineering Contradiction:

Improveaccuracy of surface mappingVSAvoidtraining time and computational resources

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The neural network is trained in advance using a comprehensive training set that covers diverse surface features and friction coefficients before actual operation. This preliminary training allows the network to store learned relationships between moment/force data and object characteristics, enabling fast inference during actual surface mapping without requiring real-time complex computations

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

Instead of performing complex real-time calculations during surface mapping, the system uses a pre-trained neural network model that has copied and stored the relationships between sensor data and object properties. This allows rapid prediction of surface topography and compliance from simple moment and force measurements

Inventive Principle:

Principle #26Copying

3Loss of information

If the whisker array contacts the object surface, then tactile information is obtained, but friction variations and surface complexity make interpretation difficult

Engineering Contradiction:

Improveinformation from tactile sensorsVSAvoiddifficulty of interpreting sensory data

Core Design Contradiction:

Loss of informationVSDifficulty of detecting and measuring

Solution Approach 1:

The system replaces direct mechanical interpretation of friction and surface contact with a neural network-based information processing approach. The network learns to compensate for friction variations and surface complexity by mapping moment and force measurements to accurate surface representations, effectively substituting complex mechanical analysis with learned computational models

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS9092737B2Systems, methods, and apparatus for 3-D surface mapping, compliance mapping, and spatial registration with an array of cantilevered tactile hair or whisker sensors

Publication Date: 2015.07.28 NORTHWESTERN UNIV

AI summary

Systems, methods, and apparatus are provided using signals from a set of tactile sensors mounted on a surface to determine a surface topography. An example method includes receiving a set of moment and force input data from one or more identified topographies. The example method includes using a neural network to receive input from a training data set based on the first set of moment and force input data from the one or more identified topographies. Network weights to be used by the neural network to produce the training data set are modified via an evolutionary algorithm that tests vectors of candidate network weights. The example method includes receiving a moment and force input from a test object surface and reconstructing the surface topology based on the neural network outputs.