MINDS CDT doctoral researcher
Using mechanical resonators to process information
Lianghong Ye is investigating whether tiny mechanical devices that vibrate, known as resonators, can help to process signals that change over time. His research explores physical reservoir computing, a brain-inspired approach that could support lower-power computing for sensors and edge AI.
Research profile: Lianghong Ye on GitHub
Research at a glance
| Research focus | Physical reservoir computing |
| MINDS theme | Nanoelectronic Technologies for AI |
| Technologies | Coupled MEMS and NEMS resonators, nonlinear dynamics, numerical modelling and device simulation |
| Potential applications | Lower-power signal processing for sensors and edge AI |
| Current status | Doctoral researcher |
| Outputs | Conference and workshop posters, 2025 |
Can a vibrating device help to compute?
Many of the signals that sensors capture change over time, such as vibrations, sound, movement or environmental readings. Making sense of this kind of sequential information with machine learning usually means running a software model on conventional digital hardware, which can require substantial computing resources and energy.
For small sensors and edge devices, sending every signal on to separate computing hardware may be inefficient or impractical.
Lianghong Ye is investigating a different route: a brain-inspired approach known as physical reservoir computing. His research examines whether coupled mechanical resonators can carry out part of the information processing through their own physical behaviour.
A resonator is a physical structure that vibrates, or responds, in characteristic ways when it is stimulated. When several resonators are coupled together, the movement of one affects the others, and their combined response to an incoming signal can be rich and nonlinear. Physical reservoir computing aims to use this changing behaviour as part of the computation, leaving a simpler trained stage to interpret the result.
Why this research matters
Processing signals where they are created
Physical reservoir computing could support ways of processing signals close to the sensor that generates them, reducing reliance on sending raw data to separate computing hardware.
Exploring lower-power computing
Using the behaviour of a physical device for part of the computation may offer a more energy-efficient way to handle some time-dependent tasks. Whether and when this holds in practice is one of the questions the research investigates.
Sensing and computing together
In the longer term, the approach may have potential applications in intelligent sensors and hardware-based AI, where a mechanical device could contribute to computing as well as sensing.
Research approach
Lianghong uses numerical modelling and device simulation to study how coupled MEMS and NEMS resonators respond to time-dependent input signals, and how their nonlinear dynamics could be used for reservoir computing tasks.
The work described on this page is based on modelling and simulation. Applications in intelligent sensors and edge AI remain longer-term possibilities that further research would need to establish.
Key terms explained
- MEMS (micro-electromechanical systems): micro-scale mechanical devices, such as tiny beams or membranes, that can move or vibrate.
- NEMS (nano-electromechanical systems): similar structures engineered at an even smaller, nano scale.
- Resonator: a physical structure that vibrates or responds in characteristic ways when it is stimulated.
- Nonlinear dynamics: behaviour in which a system’s response is not simply proportional to its input, which allows it to transform signals in complex ways.
- Reservoir computing: a machine-learning approach in which a fixed dynamic system transforms incoming signals, and only a simple final stage is trained to read the result.
- Physical reservoir computing: reservoir computing in which a physical system, such as a set of coupled resonators, acts as that dynamic system.
- Edge computing: processing data on or near the device that collects it.
Research keywords: Physical reservoir computing, MEMS, NEMS, coupled resonators, nonlinear dynamics, time-dependent data, numerical modelling, edge AI
Research and outputs
Conference and workshop posters
Lianghong presented his research as posters at two events in 2025:
- Poster, 51st International Micro and Nano Engineering Conference (MNE 2025)
- Poster, International Workshop on Reliable and Sustainable Neuromorphic Hardware (2025)
Code
Lianghong shares code through his GitHub profile.
Related MINDS research
Reservoir computing in hardware is also explored elsewhere in MINDS. Xiyue Fan has contributed to published work on memristor devices for reservoir computing. Explore Xiyue Fan’s research
Connect with Lianghong’s research
Interested in physical reservoir computing, MEMS and NEMS devices or lower-power edge AI? Contact the MINDS team about a research or collaboration enquiry, or explore Lianghong’s code on GitHub.
Not sure whether this is the right research area? The MINDS team can help direct your enquiry.