Overview
Our research in edge computation is dedicated to enabling machine learning on microcontroller and FPGA platforms. We focus on designing efficient algorithms and hardware-software co-design strategies that make it feasible to deploy intelligent systems on low-power, resource-constrained edge devices.
By bringing computation closer to the source of data, we reduce latency, enhance privacy, and minimize bandwidth usage—critical for real-time applications in domains such as acoustic sensing, radar-based activity recognition, and gesture control.
Our platforms such as RAMAN and tinyRadar demonstrate how targeted optimizations, including multiplierless computing and compact neural network architectures, can bring advanced ML capabilities to always-on systems in the field. These efforts pave the way for scalable, cost-effective, and responsive edge intelligence across a wide range of practical applications.
Demos
The following videos demonstrate various prototypes and applications built on our edge computing platforms. These include:
- Demo 1: Human Activity Recognition using mmwave radar
- Demo 2: Gesture recognition using mmWave radar on the tinyRadar edge system.
- Demo 3: Bird Hotspots Acoustic Classification System for Ecological Insights
Selected Publications
- Krishna, A., Nudurupati, S. R., Chandana, D. G., Dwivedi, P., van Schaik, A., Mehendale, M., & Thakur, C. S. (2024). Raman: A re-configurable and sparse tinyML accelerator for inference on edge. IEEE Internet of Things Journal.
- Krishna, A., Shankaranarayanan, H., Oleti, H. P., Chauhan, A., van Schaik, A., Mehendale, M., & Thakur, C. S. (2023, November). TinyML Acoustic Classification using RAMAN Accelerator and Neuromorphic Cochlea. In 2023 IEEE Asia Pacific Conference On Postgraduate Research In Microelectronics And Electronics (PRIMEAsia) (pp. 44-45). IEEE.
- Yadav, S. S., Anand, S., Nikitha, D. S., & Thakur, C. S. (2024, May). tinyradar: Lstm-based real-time multi-target human activity recognition for edge computing. In 2024 IEEE International Symposium on Circuits and Systems (ISCAS) (pp. 1-5). IEEE.
- Nair, A. R., Nath, P. K., Chakrabartty, S., & Thakur, C. S. (2024, January). Multiplierless in-filter computing for tinyML platforms. In 2024 37th International Conference on VLSI Design and 2024 23rd International Conference on Embedded Systems (VLSID) (pp. 192-197). IEEE.
- Yadav, S. S., Singh Thakur, C., MD, A., Anand, S., Munasala, M., & Kankipati, D. (2023, October). Live Demonstration: Real-time Gesture Recognition Using tinyRadar for Edge Computing. In Proceedings of the Third International Conference on AI-ML Systems (pp. 1-2).
- Kankipati, D., Munasala, M., Nikitha, D. S., Yadav, S. S., Rao, S., & Thakur, C. S. (2023, November). tinyRadar for Gesture Recognition: A Low-power System for Edge Computing. In 2023 IEEE Asia Pacific Conference on Circuits and Systems (APCCAS) (pp. 75-79). IEEE.
- Sabbella, H. R., Nair, A. R., Gumme, V., Yadav, S. S., Chakrabartty, S., & Thakur, C. S. (2022, May).An always-on tinyml acoustic classifier for ecological applications. In 2022 IEEE International Symposium on Circuits and Systems (ISCAS) (pp. 2393-2396). IEEE.
- Yadav, S. S., Agarwal, R., Bharath, K., Rao, S., & Thakur, C. S. (2022, May). TinyRadar: MmWave radar based human activity classification for edge computing. In 2022 IEEE International Symposium on Circuits and Systems (ISCAS) (pp. 2414-2417). IEEE.