Decentralized AI to Optimize Battery Performance, Lifetime, and Safety – Cognivity AI Sweden AB and Lucid Motors Inc.
| Reference number | |
| Coordinator | Cognivity AI Sweden AB |
| Funding from Vinnova | SEK 750 000 |
| Project duration | November 2025 - September 2026 |
| Status | Completed |
| Venture | Future Mobility |
Important results from the project
Yes, the main goal was achieved. The goal was to reduce the error in the battery health estimate to approximately ±2%. In a simulated fleet of 100 vehicles, the TrinityBMS achieved an error of 1.0 percentage points, compared to 3.7 for a common fleet model. The project also contributed to a toolchain that makes the models run on embedded BMS hardware and to 10–40 times faster validation. The next step is validation in SiL, HiL and on real battery packs.
Expected long term effects
More accurate estimates of battery health extend battery life, facilitate second life and reduce the need for raw materials and CO2 emissions. Smaller safety margins provide more range and lower costs. Learning takes place in the vehicle, so raw data never leaves the car. The technology can also be used in heavy vehicles, energy storage and light electric vehicles. The goal is to license TrinityBMS to vehicle manufacturers within 3–5 years, which strengthens Swedish competitiveness in software.
Approach and implementation
The project followed the plan with four work packages: requirements and data, model development, simulated validation and business development. The activities were right, but the schedule slipped. Due to a reorganization at the OEM partner, the LOI was only signed in March and data access took longer. The fleet was therefore emulated from real vehicle data, and not all parts were completed. Validation took unexpectedly long, so we made it 10–40 times faster. The collaboration worked very well.