Near-Sensor Analog Tactile Computing With Integrated Memristor Arrays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional artificial skin systems for wearable devices and robotics rely on front-end electronics for data processing, leading to high latency and energy consumption, and integrating tactile sensors with memristor arrays poses challenges that impair vector-matrix multiplication operations.
Innovation Solution
A tactile near-sensor analogue computing system that seamlessly integrates a tactile sensor array with a flexible memristive computing array, allowing direct processing of analogue sensory signals without conversion circuits, using a pyramidal pressure sensor array and a flexible Au/TiW/HfO2/Au memristor array for ultrafast and energy-efficient computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If front-end electronics are used to process tactile signals, then data processing capability is provided, but response latency increases to milliseconds and energy consumption increases to milliwatts
Solution Approach 1:
The patent merges the tactile sensor array directly with the memristor computing array, eliminating separate front-end electronics. Each tactile sensing device is directly connected to a memristive device, allowing sensing and computing to occur in the same physical location, thus reducing response latency from milliseconds to microseconds while maintaining data processing capability
Solution Approach 2:
The patent introduces an interface layer that directly couples sensor outputs to memristor inputs without requiring analogue-to-digital conversion. This intermediary connection allows analogue sensory signals to be processed directly by the memristor array's resistive network, bypassing the slow conversion process and achieving ultrafast response times
2Productivity
If front-end electronics are used to process tactile signals, then data processing is enabled, but energy consumption increases to milliwatts
Solution Approach 1:
By combining sensing and computing functions into a single integrated array, the patent eliminates energy-consuming intermediate stages such as signal amplification, analogue-to-digital conversion, and data transmission to external processors. The memristor array performs computing operations using only the charge required to modulate its conductance states, reducing energy consumption from milliwatts to microwatts
Solution Approach 2:
The memristor array is self-sufficient in processing tactile signals without requiring external power-intensive electronics. The sensing devices generate signals that directly modulate the memristor conductance through their electrical connection, allowing the system to perform computing operations using the inherent electrical properties of the materials rather than requiring additional active components
3Loss of time
If tactile sensors are integrated with memristor arrays for near-sensor computing, then response latency is reduced, but current convergence of the resistive network is affected and VMM computation is impaired
Solution Approach 1:
The patent applies different functional characteristics to different parts of the integrated array. The tactile sensing devices are optimized for signal generation, while the memristive devices are optimized for computing operations. This local differentiation allows each component to perform its function optimally without interfering with the other, maintaining VMM computation accuracy while achieving fast response times
Solution Approach 2:
The patent introduces an interface layer with specific electrical characteristics that mediates between the sensor and memristor. This interface is designed to preserve the analogue signal integrity from the sensor while providing the appropriate electrical conditions for the memristor computing operations, thus preventing degradation of VMM computation accuracy
4Productivity
If analogue signals are converted to digital format before computation, then data processing is enabled, but response speed is limited
Solution Approach 1:
The patent replaces the mechanical/electronic analogue-to-digital conversion process with a direct analogue computing approach using the memristor array's resistive network. The analogue tactile signals directly modulate the memristor conductance, and the computing operations are performed on the analogue signals themselves through electrical circuit operations, eliminating the speed-limiting conversion step while maintaining full data processing capability
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
The system achieves ultrafast response times of 400 ns with low power consumption, enabling real-time sensing and computing tasks like noise reduction and edge detection, consuming 1000 times less power than conventional systems, and facilitating advanced human-machine interactions.
Implementation Method 1
a tactile sensor array and a memristive computing array. The tactile sensor array is configured to capture data and includes a plurality of tactile sensing devices
Implementation Method 2
memristor devices, which can directly compute analogue data through physical resistive networks, are capable of solving complex tasks
Data Source
AI summary
A tactile near-sensor analogue computing system and an artificial skin system are provided. The tactile near-sensor analogue computing system includes a tactile sensor array and a memristive computing array. The tactile sensor array is configured to capture data and includes a plurality of tactile sensing devices. The memristive computing array is configured to process the data and includes a plurality of memristive devices, each of the plurality of tactile sensing devices connected to one of the plurality of memristive devices.