Session
Weekday Session 11: Advanced Technologies 2
Location
Utah State University, Logan, UT
Abstract
Accurate attitude determination ensures precise payload sensor alignment and optimal solar panel positioning, enhancing data collection and power efficiency while ensuring stable orbit maintenance. The most accurate method of attitude determination is star tracking. Star trackers collect an image of a star field and utilize a catalog of known star positions to calculate the orientation of the system relative to the stars. The errors present in this method of attitude determination tend to be on the order of ten arcseconds or less. Using traditional collection techniques, attitude estimates can be provided at update rates of 1 to 10 Hz and angular velocities are typically limited to a few degrees per second.
Neuromorphic cameras, also known as event cameras, detect changes in the input visible signal on a per pixel basis, enabling microsecond temporal resolution and significantly lower data volume. This collection method is also capable of operating with a significantly reduced power usage ( < ~0.1W) and mass ( < 50g) compared to traditional high-speed cameras. These characteristics make neuromorphic cameras ideal candidates for autonomous navigation of small satellites that commonly have stricter size, weight, and power requirements. The application of neuromorphic cameras to autonomous navigation and localization is an active area of research. Many techniques leverage this unique dataset, such as Monte Carlo Localization (MCL).
MCL is a probabilistic technique used in robotics and autonomous navigation. It employs random sampling to estimate a system's position and orientation based on a known map and sensor data. Particle filters implement MCL by creating a dynamic model of potential states through a series of particles, which are iteratively updated and reweighted to align with sensor observations and refine the system's state estimate.
ExoAnalytic Solutions utilizes a particle filter to process event camera measurements generated by the stars, leveraging MCL to determine the attitude of the satellite. A star catalog is used as the known map that the particle filter compares measurements to. This process streamlines star identification and makes it continuous which saves time compared to traditional approaches like pattern matching algorithms. The high update rate of the neuromorphic star tracker can aid autonomous navigation of small satellites by providing improved stabilization, maneuver/station-keeping efficiency, and maneuverability. In the increasingly crowded space environment, highly accurate and responsive attitude determination technology is critical to ensuring a safe space environment for all operators. This paper presents the results of a MCL particle filter applied to simulated measurements from a neuromorphic star tracker.
SSC24-XI-04-Presentation
Neuromorphic Star Tracking Using Monte Carlo Localization
Utah State University, Logan, UT
Accurate attitude determination ensures precise payload sensor alignment and optimal solar panel positioning, enhancing data collection and power efficiency while ensuring stable orbit maintenance. The most accurate method of attitude determination is star tracking. Star trackers collect an image of a star field and utilize a catalog of known star positions to calculate the orientation of the system relative to the stars. The errors present in this method of attitude determination tend to be on the order of ten arcseconds or less. Using traditional collection techniques, attitude estimates can be provided at update rates of 1 to 10 Hz and angular velocities are typically limited to a few degrees per second.
Neuromorphic cameras, also known as event cameras, detect changes in the input visible signal on a per pixel basis, enabling microsecond temporal resolution and significantly lower data volume. This collection method is also capable of operating with a significantly reduced power usage ( < ~0.1W) and mass ( < 50g) compared to traditional high-speed cameras. These characteristics make neuromorphic cameras ideal candidates for autonomous navigation of small satellites that commonly have stricter size, weight, and power requirements. The application of neuromorphic cameras to autonomous navigation and localization is an active area of research. Many techniques leverage this unique dataset, such as Monte Carlo Localization (MCL).
MCL is a probabilistic technique used in robotics and autonomous navigation. It employs random sampling to estimate a system's position and orientation based on a known map and sensor data. Particle filters implement MCL by creating a dynamic model of potential states through a series of particles, which are iteratively updated and reweighted to align with sensor observations and refine the system's state estimate.
ExoAnalytic Solutions utilizes a particle filter to process event camera measurements generated by the stars, leveraging MCL to determine the attitude of the satellite. A star catalog is used as the known map that the particle filter compares measurements to. This process streamlines star identification and makes it continuous which saves time compared to traditional approaches like pattern matching algorithms. The high update rate of the neuromorphic star tracker can aid autonomous navigation of small satellites by providing improved stabilization, maneuver/station-keeping efficiency, and maneuverability. In the increasingly crowded space environment, highly accurate and responsive attitude determination technology is critical to ensuring a safe space environment for all operators. This paper presents the results of a MCL particle filter applied to simulated measurements from a neuromorphic star tracker.
