- What is your contribution to the project (what’s your project part about)?
Our role in the project is to develop a new generation of brain-like AI algorithms and microcomputing architectures to process continuous multimodal sensor measurements at the edge. By mimicking the processing capabilities of our nervous system, we target to boost energy and computational autonomy so as to enable customizable, early assessment of (a)biotic water pollution with high accuracy, wherever and whenever needed.
- What’s the benefit/opportunity for society from the BioSensei research project?
We expect our systems to have profound impact on achieving a comprehensive and efficient monitoring of aquatic environments, extending the lifespan of the biosensors with minimum calibration requirements, and unlocking access to crucial quantitative and qualitative information measured inline. All to prevent contamination and to allow a more sustainable management of water resources, making the system compatible with the severe energy constraints of remote or mobile operation.
- What are the scientific advancements you most look forward to in this field of research?
We are looking forward on developing new methods at the intersection of deep learning and neuroscience so as to overcome current challenges limiting water analysis in real settings: novel approaches to quickly self-calibrate to sensor variability in complex and changing environmental conditions, as well as dynamical neuron models and networks to efficiently integrate the temporal sensor readings in situ. Besides curtailing energy consumption and hardware requirements, these algorithms should advance to require smaller datasets and to provide interpretable insights for predictive analytics or personalized remediation.
The image shows the spiking response of a biorealistic neural network forecasting a temporal signal, and depicts its implementation on-chip
Picture of the team From left to right: Ferran Delga, Cecilia Jimenez, Josep Maria Margarit, Alex Fulleda, Carlos Álvarez