Modified Nonlinear Hysteresis Approach for a Tactile Sensor

Area of Science:

• Materials Science
• Robotics
• Artificial Intelligence

Background:

• Soft tactile sensors utilizing piezoresistive materials are crucial for large-area sensing applications.
• Hysteresis in these sensors significantly impacts their operational accuracy, presenting a major challenge.

Purpose of the Study:

• To develop and validate a novel approach using a backpropagation (BP) neural network to mitigate hysteresis nonlinearity in conductive fiber-based tactile sensors.
• To enhance the accuracy and reliability of soft tactile sensors for advanced applications.

Main Methods:

• Designed four sensor units based on conductive fibers to collect output resistance data under varying force sequences.
• Trained a backpropagation neural network using the collected data to correct resistance values and minimize prediction errors.
• Conducted validation experiments to assess the effectiveness of the BP network in reducing hysteresis error.

Main Results:

• The backpropagation network training demonstrated excellent convergence, effectively adjusting parameters to minimize prediction errors.
• The proposed method successfully reduced the maximum hysteresis error from 24.2% to 13.5% of the sensor's full-scale output.
• The trained BP network accurately predicted sensor output resistances, validating the approach's efficacy.

Conclusions:

• The BP neural network approach offers a promising solution for enhancing the accuracy of piezoresistive soft tactile sensors by mitigating hysteresis nonlinearity.
• This advancement improves the reliability and efficiency of tactile sensors, expanding their utility in soft robotics, wearable technology, medical devices, and consumer electronics.
• While complete hysteresis elimination may be unachievable, this method effectively modifies nonlinearity, leading to superior sensor output accuracy.