Tracking the Dots in Orbit
summary
A fully autonomous telescope called the aware station is built in Western Australia and goes operational in 2023 to address ~45,000 regularly tracked objects while millions remain unseen; software is mostly Python, containerized, runs local processing at about 1 Hz on a Tesla T4, and uses a YOLO v8 model trained on ~300 annotated images. Remote power cycling, udev device rules, simulators for end-to-end testing, and CDM outputs prove the system works in practice and scales toward networked stations.
AI-written summary · about
chapters (27)
- 0:00Space safety and tracking intro
- 1:48Why tracking orbit matters
- 5:59Collision avoidance challenges
- 8:08Position uncertainty with telescopes
- 9:34Accountability in space
- 10:59Kessler syndrome risks
- 12:01Transparency and law enforcement
- 13:47Satellite failures from Earth
- 14:30Space debris tracking requirements
- 15:55AWARE Station project intro
- 17:22Automated telescope design
- 20:36Handling remoteness and failures
- 22:23udev rules deep dive
- 24:56Remote access and network reliability
- 27:03Single points of failure
- 28:26Crash recovery and stateless design
- 29:08Developing with simulators
- 30:10Simulating USB on Linux
- 32:39End-to-end hardware testing
- 34:47Adapting simulators to hardware
- 35:28Real-time image processing
- 37:16Detections into cohesive tracks
- 37:58Log collection importance
- 39:02Automating with weather data
- 40:29Space traffic management future
- 42:15Collaboration in orbit
- 43:02Q&A on satellite maneuvering
related talks
57mDrones & AI - What's all the buzz about ?
22mThe Untold Story of Space
01h 00mA Future City
21mAI, Embedded System and Robotics
10mPositional tracking and new spatial interactions at AWE 2015
08mAR for Flight Control at AWE 2015
01h 02mAutonomous Exploration Robotics with a Raspberry Pi
21mDesigning the product that can do anything