Bestowing the human touch upon robots

Creating an E-skin that can rival its human counterpart has long been a challenge as replicating multimodal tactile perception in a single, compact device is very complex. Existing devices often integrate multiple sensors to detect various stimuli, such as temperature, pressure and texture — leading to overly complicated designs, signal inaccuracies and exorbitant costs.

Simplicity is what robots are after. The NUS team’s E-Skin is composed of just two advanced sensors: the transient voltage artificial neuron (TVAN) and the sustained potential artificial neuron (SPAN).

The TVAN detects dynamic tactile stimuli, such as vibration, texture and material, using a triboelectric-based mechanism that generates electrical signals in response to mechanical interactions with surfaces. The sensor takes the form of a honeycomb, which makes it more sensitive to different textures and materials — much like how our fingerprints can tease out subtle surface changes.

The SPAN, on the other hand, is an ionic hydrogel-based sensor designed to measure static tactile information, such as pressure and temperature. Unlike conventional sensors, the SPAN generates its own voltage without requiring an external source, reducing energy consumption and simplifying its design, making it suitable for applications that require continuous monitoring.

“By integrating the TVAN and SPAN into a single unit, our E-Skin converts detected physical stimuli into electrical signals, while machine-learning algorithms make sense of these signals to determine an object’s texture, roughness, hardness, material and temperature,” says Assoc Prof Lee. “Machine learning also helps mitigate the challenges of signal interference and variability in real-world applications, compensating for inconsistencies in applied force, speed and environmental conditions.”