In this research area we look for solutions that enable the implementation of AI in IoT and wearable devices, which are typically constrained in terms of computing power, memory and energy. We tackle the topic at three levels of abstraction.
Firstly, we look into the demands of the applications and how they can be accommodated into existing ultra-low power embedded platforms.
Secondly, we look for new ways to efficiently integrate multiple sensors in a platform for our multimodal biomedical applications.
Finally, we analyze the applications to identify their computational hotspots (kernels) and propose novel hardware architectures and accelerators to compute them more efficiently.
ML-enabled IoT devices and embedded AI
Development of new cost-effective and modular hardware platforms for AI-enabled IoT devices
Novel ULP MPSoC and hardware accelerator architectures for energy-efficient execution of biomedical applications
Personalized Detection of Epileptic Seizure in the Internet of Things (IoT) Era
Implementation of biomedical applications on ULP embedded platforms