MoDeVVa 2019 Program

Session: Keynote

Iulia-Elena Dragomir

Iulia Dragomir is a system/software engineer at GMV Aerospace and Defence in the On-Board Autonomy Division. She obtained her PhD in Software Engineering and Formal Methods from University of Toulouse in 2014. During her postdoctoral studies, she has worked at Aalto University, Finland, on the RCRS framework (http://rcrs.cs.aalto.fi/), and at Verimag – Universite Grenoble Alpes, France, in the frame of H2020 ESROCOS and ERGO projects on the BIP tools (http://www-verimag.imag.fr/RSD-Tools.html?lang=en).

Her current work focuses on the development of autonomous systems, from system design to implementation. In particular, in the H2020 ADE project (https://www.h2020-ade.eu/), her activities cover the coordination and development of autonomous planetary robotics inspired by the Mars Sample Return mission including autonomous on-board software, ground control station and system simulator.

Designing autonomous robotics systems: from models to functionally correct implementations

The PERASPERA Programme (https://www.h2020-peraspera.eu/) aims to develop key space robotics technologies at a significant scale and technology readiness level suitable for future commercial exploitation. The first call of the programme, finished beginning of 2019, covered the common basic building blocks ranging from robotics component library to autonomy (ERGO - https://www.h2020-ergo.eu/), data fusion, integrated sensors, and payloads. The second call, currently on-going, develops multiple mission demonstrators using the common building blocks, among which ADE (https://www.h2020-ade.eu/) targets autonomous exploration of terrains involving long traverses and scientific detection.

In this talk we present the challenges and results obtained in ERGO. The ERGO project has developed a framework for autonomy that allows commanding a (space) robotic system via high-level goals. The autonomy capabilities are implemented by multiple components: on-board mission planning for dynamically generating plans from high-level goals, scientific detection of targets of interest from images, rover navigation and robotic arm motion planning and execution, and fault detection, isolation and recovery (FDIR). For this, the framework embodies Artificial Intelligence technologies, as well as formal correct-by-construction component synthesis.

The capabilities of the framework have been tested in two scenarios. The first one involved the in-orbit autonomous servicing of a damaged spacecraft, tested in the dynamic simulation facility at GMV. The second scenario, inspired from the Mars Sample Return, achieved an autonomous traverse in the Moroccan desert of 1.4km in 8h.

Developing such complex systems raises many challenges at every step of the process, mainly focused on: approach for system design, tools and formalisms for system design, integration of different computational models, scalability of verification and validation. To cope with such difficulties, the framework implements a rigorous model-based development approach based on the TASTE model-driven architecture tool (developed by ESA) and the BIP tools for formal verification and validation (developed by VERIMAG). Yet, the observations from the development and field testing phases showed a gap between the needs of industry in terms of technologies and tools, and their current capabilities.

In this talk we describe the approach, techniques and technologies put in place in ERGO for system design and development, the experience of using them and the lessons learned. We present the achievements of the field trials of the two scenarios. We conclude with the work on-going in ADE to improve and enhance the framework capabilities, and to make it ready for future missions.