Decentralized AI to Optimize Battery Performance, Lifetime, and Safety
This collaboration between Cognivity AI Sweden AB and California-based Luxury EV Manufacturer explores the transition from static, rule-based battery management software to adaptive, AI-driven models deployed directly on the battery system.
Electric vehicle (EV) manufacturers currently face challenges in accurate battery diagnostics, leading to discrepancies in estimated range, charging times, and State-of-Charge (SoC) readings. Current Battery Management System (BMS) software is largely based on static lab data applied generically across all batteries, which ignores real-world variability and individual driving behaviors. Cloud-based alternatives offer limited improvements due to low data resolution, high processing costs, and strict data privacy regulations such as GDPR.
Consequently, these diagnostic limitations often result in oversized battery packs, range anxiety, and costly inefficiencies. Addressing these challenges requires a decentralized approach that can provide real-time, personalized diagnostics without the need for raw data transfer, thereby extending battery lifetime and ensuring data security.
The Project
The project investigates how decentralized artificial intelligence can be integrated into existing battery management systems. By tailoring proven AI models from stationary energy storage to automotive environments, the focus lies on validating the system’s accuracy, robustness, and adaptability in three key areas:
- Decentralized Architecture and Data Security: The study analyzes the implementation of AI directly on the battery system. This architecture ensures GDPR compliance by allowing vehicles to exchange learned insights rather than raw data, creating a secure foundation for fleet-level intelligence.
- Diagnostic Accuracy and Optimization: Using laboratory data and engineering fleet data, the project refines models for dynamic automotive conditions. The objective is to rigorously evaluate the models' ability to reduce State-of-Charge (SoC) and State-of-Health (SoH) estimation inaccuracies from ±30% to approximately ±2%.
- System Integration and Cross-Platform Learning: The project evaluates how the multi-layer machine learning architecture can be seamlessly integrated using standard measurements (pack current, cell voltages, temperatures) without requiring additional sensor costs, enabling continuous learning across different climates and usage profiles.
This collaboration between Swedish and American experts bridges the gap between advanced predictive analytics and automotive engineering. The results will provide EV manufacturers with a validated foundation for integrating AI-driven solutions into real vehicles, ultimately supporting a scalable transition toward smarter, shared, and more sustainable transport systems.
"By collaborating across borders with a California-based luxury EV manufacturer, we are able to combine Sweden’s strengths in AI and predictive analytics with the fast-moving automotive innovation ecosystem in California."
DR. CHRISTIAN FLEISCHER, Co-Founder/CEO, Cognivity
The project focuses on the development of a hierarchical, cloud-supported (yet not cloud-dependent) AI software stack for electric mobility. The system consists of three integrated layers designed to enable real-time vehicle adaptation, shared learning across similar battery types, and continuous fleet-wide improvements without compromising privacy or performance.
The software stack operates through the following three layers:
- Local AI Integration: AI models operate directly on each vehicle, facilitating real-time, on-board optimization tailored to specific driving behaviors, usage patterns, and individual battery conditions.
- Battery Clustering: Similar battery profiles are grouped across the fleet. This allows for shared insights and coordinated learning while keeping raw data securely localized.
- Fleet-Wide Learning: The system continuously improves through aggregated learning. Software updates are systematically deployed to individual vehicles to maximize range, increase uptime, and extend overall battery lifespan.
"By collaborating across borders with a California-based luxury EV manufacturer, we are able to combine Sweden’s strengths in AI and predictive analytics with the fast-moving automotive innovation ecosystem in California. One of the biggest advantages is the ability to work closely with real engineering challenges from the vehicle side, access relevant data quickly, and validate ideas in an environment where new technology can move from concept to testing very rapidly.
This kind of collaboration creates a much shorter path between research and real-world deployment. It allows us to train and refine our models on practical use-cases, while also solving challenges around secure international data collaboration and privacy-preserving AI. The result is not only better battery diagnostics and performance, but also a stronger foundation for the next generation of intelligent, adaptive battery systems."
– Christian Fleischer, CEO, Cognivity AI.
Key Deliverables
- Requirements and Architecture: A defined system architecture and a secure, compliant data pipeline for AI-driven BMS.
- AI Model Prototypes: Development and training of multi-layer machine learning models for SoC, SoH, and Remaining Useful Life (RUL) estimation.
- Simulated Validation: Benchmarking of AI performance against current in-vehicle and cloud-based diagnostics using simulated driving cycles and historical fleet data.
- Strategic Expansion: Preparation for future Hardware-in-the-Loop (HiL) validation and the establishment of further collaborations within the US EV ecosystem.
Partners: Cognivity AI Sweden AB and California-based Luxury EV Manufacturer
Project Manager: DR. Christian Fleischer, Cognivity AI Sweden AB
Period: 2025 – 2026
This collaboration is the result of a joint effort led by Future Mobility, co-funded by Vinnova and the Swedish Energy Agency. The initiative is designed to foster partnerships and strengthen the research and innovation (R&I) network between Sweden and the United States to speed up the development of sustainable mobility.
Do you want to explore more?
Hans Pohl
