Duration: 36 months
Seeing in space is nothing like seeing on Earth. Extreme lighting variations, rapid motion, and deep shadows challenge traditional cameras, causing motion blur, limited dynamic range, and limitations in latency and bandwidth. Event-based cameras offer an alternative with 120dB dynamic range, microsecond latency, and milliwatt power consumption, but lack absolute intensity information, requiring fusion with frame-based cameras for robust space applications. However, efficiently combining these modalities on resource-constrained, space-grade hardware while ensuring low-latency and energy-efficient processing is non-trivial. This project aims to address these challenges by developing computationally efficient, low-power methods to combine event and frame data on space-qualified hardware. Tracking serves as the core focus, as it is fundamental to various space tasks, with visual odometry (VO) as one of its most critical applications. By improving tracking, our work enhances VO systems for planetary landing, real-time monitoring for space situational awareness, and motion estimation for rendezvous and docking, all within the strict computational and power constraints of space systems. We plan a collaborative partnership with ESA's ZTH Robotics at ESTEC, leveraging the GRALS testbed to benchmark our system under realistic space-like conditions. Additionally, we will work with Thales Alenia Space, drawing on their expertise in deploying space-qualified solutions as prime contractor for ESA’s VISNAVMars, where quasi-real-time (10Hz) Guidance, Navigation, and Control was demonstrated on low-power space-grade hardware for Entry, Descent, and Landing. Thales will provide real-world use cases, particularly in landing and rendezvous, and contribute expertise in synthetic image modeling tools for controlled testing. These efforts will ensure that our system is optimized for the extreme constraints of space while maintaining robust performance in real missions.
