Olli Knuuttila

My research focuses on autonomous mission impact optimization of CubeSat-standard satellites sent to explore small solar system objects such as asteroids or comets. To achieve highly autonomous and productive missions, following key research areas were identified: 1) optical navigation near a target object, 2) novelty/interestingness detection in images and other types of measurements and 3) AI methods to maximize detected novelty/interestingness. The research will identify the most promising existing approaches and adjust them to our use case.

Reliable and accurate navigation with minimum support from the ground is essential as the communication link to Earth might be unstable and round trip times can be significant. Optical navigation has low requirements on the spacecraft hardware making it possible to include better mission-related instruments. With reliable navigation, it is possible to let the spacecraft do more autonomous exploration as the risk of catastrophic failure decreases. Autonomous exploration and efficient downlink usage are enabled by the capability of detecting interesting features in images or measurements.

For optical navigation, we will look into different local feature detection and extraction algorithms in combination with different scenarios for providing reference features needed for absolute and relative navigation. Traditional scheme with feature matching followed by geometric rejection will be compared to a deep learning based approach by using both synthetic and actual NavCam images taken by e.g. Rosetta spacecraft. For novelty detection, we will evaluate algorithms from e.g. ESA MASTER and NOAH projects, while also investigating the usage of Kullback-Leibler divergence and artificial curiosity as pioneered by J. Schmidhuber.

I'm confident that at least some algorithms validated by simulations can be tested in space, as I have already done some work for Aalto-1 and Suomi100 CubeSats. Additionally, I'm responsible on behalf of Aalto University for the optical navigation in the ESA funded binary-asteroid mission APEX, which is a CubeSat that will be deployed from the HERA mothership after it has arrived at 65803 Didymos.