Scale-Covariant and Scale-Invariant Gaussian Derivative Networks

This paper presents a hybrid approach between scale-space theory and deep learning, where a deep learning architecture is constructed by coupling parameterized scale-space operations in cascade. By sharing the learnt parameters between multiple scale channels, and by using the transformation properties of the scale-space primitives under scaling transformations, the resulting network becomes provably scale covariant. By in addition performing max pooling over the multiple scale channels, or other permutation-invariant pooling over scales, a resulting network architecture for image classification also becomes provably scale invariant. We investigate the performance of such networks on the MNIST Large Scale dataset, which contains rescaled images from the original MNIST dataset over a factor of 4 concerning training data and over a factor of 16 concerning testing data. It is demonstrated that the resulting approach allows for scale generalization, enabling good performance for classifying patterns at scales not spanned by the training data.

VHF Omnidirectional Range (VOR) Experimental Positioning for Stratospheric Vehicles

The usage of aeronautical radio-frequency navigational aids can support the future stratospheric aviation as back-up positioning systems. Although GNSS has been extensively redundant in the last years of space operations, radio NavAids can still be supportive of navigation and tracking for novel mission profiles. As an example, in 2016, VHF Omnidirectional Range (VOR) has been proven to work well above its standard service volume limit on a stratospheric balloon flight with the STRATONAV experiment. While VOR provides the “radial” measurement, i.e., the angle between the Magnetic North and the line between the receiver and the transmitting ground station, the intersection of two or more radials at a time allows to perform ground track reconstruction for the vehicle to be tracked. This paper reports the results from the data re-processing from STRATONAV: the acquired radials have been intersected in order to achieve positioning. The radials interfacing method, the position calculation methodology, and the data acquisition strategies from STRATONAV are reported together with the data analysis results.

Modeling the Effects of Machine Rigidities on Joint Work Strategies

One of the challenges in designing resilient human-machine systems is that machine capabilities are inherently rigid. A resilient joint cognitive system can anticipate and adapt to changing work demands effectively, but limitations of machines can make this adaptation constrained and less fluid. By identifying and accommodating for these rigidities in the design of human-machine system architectures, developers can build human-machine systems that support multiple contexts. This paper proposes a work-modeling approach for analyzing joint human-machine work strategies, focusing on identifying interdependencies that would support opportunistic adaptation and reduce the risk of machine rigidity leading to brittle failures of a human-machine system. The approach is applied to a case study in space operations to demonstrate how interdependencies can be identified and evaluated. The results of this analysis provide early insight into how team adaptation and machine limitations can be systematically accounted for in system architecture design.

10-Year Anniversary of the European Proximity Operations Simulator 2.0—Looking Back at Test Campaigns, Rendezvous Research and Facility Improvements

Completed in 2009, the European Proximity Operations Simulator 2.0 (EPOS 2.0) succeeded EPOS 1.0 at the German Space Operations Center (GSOC). One of the many contributions the old EPOS 1.0 facility made to spaceflight rendezvous is the verification of the Jena-Optronik laser-based sensors used by the Automated Transfer Vehicle. While EPOS 2.0 builds upon its heritage, it is a completely new design aiming at considerably more complex rendezvous scenarios. During the last ten years, GSOC’s On-Orbit-Servicing & Autonomy group, who operates, maintains and evolves EPOS 2.0, has made numerous contributions to the field of uncooperative rendezvous, using EPOS as its primary tool. After general research in optical navigation in the early 2010s, the OOS group took a leading role in the DLR project “On-Orbit-Servicing End-to-End Simulation” in 2014. EPOS 2.0 served as the hardware in the loop simulator of the rendezvous phase and contributed substantially to the project’s remarkable success. Over the years, E2E has revealed demanding requirements, leading to numerous facility improvements and extensions. In addition to the OOS group’s research work, numerous and diverse open-loop test campaigns for industry and internal (DLR) customers have shaped the capabilities of EPOS 2.0 significantly.