Imagine if computers could learn and think more like a human brain instead of needing massive, power-hungry data centers. This project builds AI that uses very little electricity and can adapt to new situations on the fly. It's like moving from a giant industrial furnace to a smart, efficient home heater that only uses energy when and where it's needed.
Current AI requires massive amounts of energy and expensive annotated data, making it too costly and carbon-heavy for real-time use in resource-limited edge environments.
A system of brain-like neural networks and data optimisation pipelines. It includes an explainability framework to make AI decisions understandable to humans.
If you are a robotics manufacturer dealing with high energy costs and slow response times in remote factories — this project developed brain-like neural networks that provide low-power consumption and real-time constraints. This allows robots to adapt to changing conditions without needing a constant cloud connection.
If you are a smart sensor developer dealing with limited battery life in the field — this project developed neuromorphic implementations that are resource efficient. This enables complex AI processing directly on the device, reducing the need to send data to the cloud.
If you are an autonomous fleet operator dealing with unpredictable environments and unreliable connectivity — this project developed AI with better cross-task generalisation. This ensures vehicles remain operationally robust and flexible when facing novel domains.
Based on available project data, specific pricing or implementation costs are not provided; however, the project aims to reduce costs associated with assembling training data and energy consumption.
The project is designed for the edge-cloud continuum, meaning it is built to scale across different hardware platforms and computing infrastructures depending on the task scenario.
Based on available project data, the specific IP and licensing agreements are not listed, but the consortium includes 9 industrial partners and 7 SMEs.
Yes, the project integrates an explainability framework specifically to promote trust and empower the human user, aligning with trustworthy AI ambitions.
Based on available project data, the project runs from 2024-01-01 to 2027-02-28, with 14 total deliverables guiding the development and evaluation process.
The project is heavily industry-driven, with a 75% industry ratio consisting of 9 companies, 7 of which are SMEs. This strong commercial presence, coordinated by KTH Royal Institute of Technology and spanning 8 countries, suggests the results are being developed with immediate market application and industrial constraints in mind rather than just academic curiosity.
Contact KUNGLIGA TEKNISKA HOEGSKOLAN in Sweden
Contact us to connect with the 7 SMEs developing this neuromorphic AI.
