E7140 FusionSafe
| Reference number | |
| Coordinator | SafeRadar Research Sweden AB |
| Funding from Vinnova | SEK 4 931 850 |
| Project duration | March 2025 - March 2028 |
| Status | Ongoing |
| Venture | Eurostars |
Purpose and goal
The goal is to develop a deep learning model that fuses signals from multiple cameras and radar sensors of multiple objects in complex environments. The model, called FusionSafe, is part of an overall security software in IoT security systems for autonomous off-road vehicles and machines. The aim is to increase safety, productivity, and enable 24/7 operation.
Expected effects and result
The expected result is two fine-tuned versions of FusionSafe for two specific application areas: FusionSafe Agbot for agricultural robots, and FusionSafe Quarry for heavy machinery in quarrying and mining. Both require high safety performance and are challenging due to dust, rain, fog and their complex, dynamic nature. These will be demonstrated to potential customers towards the end of the project.
Planned approach and implementation
The project spans for 36 months and is divided into six different work packages with associated milestones: Project management, pre-training of AI networks for sensor fusion (TRL5), fine-tuning of AI networks for specific use cases (TRL6), demonstration and documentation (TRL7), data management, and requirements management and validation. All parties have a vested interest in expanding their knowledge in the field of automated vehicle safety systems, especially in sensor fusion and AI.