In recent years, artificial intelligence (AI) has been applied to a wide range of safety-critical domains, such as automotive, robotics, and aviation. Especially the automotive and robotics domains have seen a rapid increase in the number of AI-based systems that are being deployed in real-world applications. However, real-world applications in the aviation domain are still sparse, given the challenges of AI engineering in combination with strict safety requirements. A first possible application of AI in the aviation domain might be the future collision avoidance system Airborne Collision Avoidance Systems X (ACAS X). The goal of collision avoidance systems is to issue advisories to the pilot to avoid near mid-air collisions (NMACs). The two important variants of ACAS X for this work are ACAS XA, providing vertical advisories and meant as a drop-in replacement for current systems in commercial air flight, and ACAS Xu, providing horizontal advisories for the ever-growing unmanned aircraft systems market. This work brings both variants closer to real-world deployment by implementing a vertical collision avoidance system, based upon ACAS XA, and a horizontal collision avoidance system, based upon ACAS Xu, for the research flight simulator FlightGear. Using advisories given by this implementation, this work furthermore provides an auto-avoid function that can command an airplane in FlightGear to safely avoid NMACs. Finally, this work will show that the ACAS X implementation can avoid collisions in a simulated environment. For this task, an Operational Design Domain will be defined serving as a basis for safety considerations and evaluating the implementation of the ACAS X. In the end, simulation-based testing will be used separately for VCAS and HCAS showing the successful utilization of advisory predictions as autopilot inputs. Summarizing, this work not only presents an open-source implementation of ACAS XAand ACAS Xu for FlightGear but also shows how the generated advisories can be used to successfully avoid NMACs.

A well-known challenge in rover teleoperation is the operator’s lack of situational awareness (SA). This often leads to an inaccurate perception of the rover’s status and surroundings and, consequently, to faulty decision-making by the operator. We present a novel teleoperation interface to control the locomotion of a ground rover with a 7DOF force feedback device (sigma.7), while providing haptic feedback to ensure appropriate SA. In particular, the device provides proprioceptive cues to convey the rover’s attitude. This can be particularly useful for environments with insufficient visual cues to estimate attitude (e.g., a cave). In systematic experimental trials controlling a robot in an outdoor environment, we evaluated the validity of employing sigma.7 as an alternative to a standard joystick. We tested the use of attitude as an aid to situational awareness. We found no significant detriment in manoeuvrability compared to a conventional joystick, thus validating the sigma.7 as an effective control device. Regarding SA, results showed no statistical difference between the visual and haptic cues for attitude feedback, thus validating the haptic method as an effective alternative to offloading the visual channel by conveying attitude information through the haptic channel instead of visual cues. Finally, qualitative observations of the participant’s behaviour during the experiments showed that operators with haptic feedback were comprehensively aware of the rover’s status.

Intelligent robotic coworkers are considered a valuable addition in many application areas. This applies not only to terrestrial domains, but also to the exploration of our solar system. As humankind moves toward an ever increasing presence in space, infrastructure has to be constructed and maintained on distant planets such as Mars. AI-enabled robots will play a major role in this scenario. The space agencies envisage robotic co-workers to be deployed to set-up habitats, energy, and return vessels for future human scientists. By leveraging AI planning methods, this vision has already become one step closer to reality. In the METERON SUPVIS Justin experiment, the intelligent robotic coworker Rollin’ Justin was controlled from Astronauts aboard the International Space Station (ISS) in order to maintain a Martian mock-up solar panel farm located on Earth to demonstrate the technology readiness of the developed methods. For this work, the system is demonstrated at AAAI 2019, controlling Rollin’ Justin located in Munich, Germany from Honolulu, Hawaii.

On June 5that 21:10 CEST, the European Space Agency's Telerobotics and Haptics Laboratory conducted the first in history haptic handshake between space and Earth. This demonstration was part of the Haptics-2 experiment, which makes use of two haptic joysticks; one located in space on-board the International Space Station (ISS), and the other one on ground at ESA's laboratory in the Netherlands. Both systems can be connected through the TDRSS geosynchronous satellites via a real-time data link with sufficient performance to perform haptic teleoperation tests. This paper describes the mechatronic design of the haptic devices, the software and the system implementation required for the Haptics-2 experiment. It reports results of a real-time high-frequency characterization of the satellite data-link indicating a round-trip time delay of approximately 850 milliseconds and presents mechatronic performance data of the haptic joysticks. Moreover, the paper presents experimental results from the first haptic handshake conducted between space and ground with astronauts on-board the ISS.

An expanded model inversion flight control strategy using sliding mode online learning is presented and implemented for an small unmanned aircraft system (UAS). These low-cost aircraft are very susceptible towards nonlinearities like atmospheric turbulence, model uncertainties or system failures, so adaptive flight control strategies possess a high potential to improve the degree of automation of such systems. The concept of feedback linearization is combined with feedforward neural networks to compensate for inversion errors and other nonlinearities. The backpropagation-based adaption laws of the network weights are extended with terms to take into account any approximation error and the so called e-modification. Within these adaption laws the standard gradient descent backpropagation algorithm is expanded with the concept of sliding mode control (SMC), which enables stable adaptivity of the learning rate and so tends to offer a higher speed of convergence. The definition of the sliding mode function assures convergence of the network's output error towards the sliding surface, while considering the system's stability. The SMC-based flight control strategy is tested and compared with the standard gradient descent backpropagation algorithm.

Implementing adaptive flight control strategies into unmanned aerial systems (UAS) contains a high potential to improve the degree of automation. This is especially the case regarding automatic operation under difficult atmospheric conditions or even system failures. A neural control strategy enables the UAS to improve its flight characteristics and to respond to unknown, non-linear flight conditions. Here, a learning flight control system for a fixed-wing UAS is realised using a systematic two-stage approach by firstly implementing a sustainable offline-trained basic knowledge and subsequently improving these characteristics during flight. Within the automated offline-step large groups of neural networks are trained with the required behaviour, which is derived from measured data. This phase shows that the necessary learning task can be achieved by multi-layered feedforward-networks. The training success of all networks is then evaluated with statistical methods and networks are selected for online application. The online learning step is realised with a control architecture comprising a neural network controller and a neural observer which predicts the system's dynamics and delivers the training signal for the contoller network. An important element of the control strategy is to determine a consistent error signal for the online training of the neural controller. This is done by backpropagation of a measured error through the inverse dynamics of the observer network. Since the inverse dynamics have to be very precise in order to train the controller adequately, a stable sliding mode control (SMC) algorithm for network training is introduced. This online adptive algorithm significantly improves the observer's charcteristics and with it the system's performance.

A point-to-point routing algorithm with three new features is presented. First, the router makes optimal use of oversized, rectangular contacts. Second, it allows different wire width on different layers, with the layers having complete freedom as to routing direction. These two features make the algorithm attractive for MOS layout applications. Finally it is able to realize all-angle routing and to accept all-angle obstacles, a feature interesting for hybrid and PCB routing. The router is gridless and guarantees a solution if one exists. Since it is based on computational geometry algorithms, it offers a low run-time complexity. The ideas have been implemented in a prototype version for 45° routing. The results indicate that the router performs well, even on large designs.

This paper describes a plowing procedure which moves and modifies cells and routing. The plow pushes aside all obstacles one by one, thus penetrating into the layout. The process leads to a compaction at the front side of the plow. The compaction is performed in accordance with the minimum spacing rules and with automatic jog insertion. Objects of the layout representation are cells, contacts, and wires. A wire is represented by a path of segments which lie in one of two layers, at angles of multiples of 45°. The generated blank space is used for adding new objects or resizing geometries.