In order to explore beyond the near-Earth environment and to the wider solar system, autonomous spacecraft (satellites, rovers, etc.) are required that can perform intelligent decision making and can adapt to unpredictability in the environment. This has motivated several projects into the design of robotic controllers for pinpoint landing, spacecraft rendezvous and docking, and satellite formation, typically using an Evolutionary Robotics approach or, more recently, machine learning. We also perform research into the automation of space mission design, the development of new learning algorithms, and natural language processing. Projects in these areas have led to contributions relevant to the wider AI community.
Transfer Learning for Hyperspectral Images01 Feb 2019
Transfer Learning for Hyperspectral Images01 Feb 2019 The conventional transfer learning approach to solve computer vision problems with Deep Learning relies on large generic supervised datasets (e.g. ImageNet), which are not necessarily available for non-RGB imagery (i.e. HSI, single channel medical, radar, etc. imagery). We are investigating ways to allow transferring RGB pre-trained networks ...
G&C Networks - Deep architectures for real time optimal actions01 Jan 2017 Recent research on deep learning, a set of machine learning techniques able to learn deep architectures, has shown how robotic perception and action greatly benefits from these techniques. In terms of spacecraft navigation and control system, this suggests that deep architectures may be considered now to drive all or part of ...
Differential Intelligence01 Oct 2016
Differential Intelligence01 Oct 2016 Many of the most celebrated techniques in Artificial Intelligence would not be as successful if, at some level, they were not exploiting differentiable quantities. The back-propagation algorithm, at the center of learning in artificial neural networks, leverages on the first and (sometimes) second order differentials of the error to update the n...
Optimal orderings of k-subsets for star identification15 Apr 2016 Star trackers are an important part of spacecraft to determine the attitude. The device concists of one or multiple cameras that detect stars and identify them based on the visible constellations. From these visible constellations it is possible to compute the orientation of the spacecraft. This is particularly challenging in the lost-in-space s...
Game theoretic analysis of the space debris removal dilemma30 Jun 2015 In this study we analysed space debris removal efforts from a strategic, game-theoretical perspective.
Interplanetary Trajectory Planning with Monte Carlo Tree Search01 Nov 2014 Planning an interplanetary trajectory is a very complex task, traditionally accomplished by domain experts using computer-aided design tools. Recent advances in trajectory optimization allow automation of part of the trajectory design but have yet to provide an efficient way to select promising planetary encounter sequences. In this project, we ...
Active space debris removal trajectory design01 Oct 2014 The space close to our planet is getting more and more polluted. Orbiting debris are posing an increasing threat to operational orbits and the cascading effect, known as Kessler syndrome, may result in a future where the risk of orbiting our planet at some altitudes will be unacceptable.
Novelty Search for Soft Robotic Space Exploration01 Mar 2014 The use of soft robots in future space exploration is still a far-fetched idea, but an attractive one. Soft robots are inherently compliant mechanisms that are well suited for locomotion on rough terrain as often faced in extra-planetary environments.
Recurrent Product Unit Neural Networks01 Nov 2013
Recurrent Product Unit Neural Networks01 Nov 2013 Time Series Forecasting (TSF) consists on estimating models to predict future values based on previously observed values of time series, and it can be applied to solve many real-world problems. This project is focused on Product Unit Neural Networks (PUNNs) and its application to TSF.
Astro Drone - A crowdsourcing game to improve vision algorithms17 Apr 2013 For autonomous robots to operate successfully in unknown environments, their computer vision algorithms need to generalize over many different environments. We propose crowdsourcing as a methodology for gathering large and varied robotic data sets. To this end, we seek to learn a relationship between the visual cues of an object and its distance...
Space Gaits17 Oct 2012
Space Gaits17 Oct 2012 Here we study the relation between walking gaits and environmental conditions such as gravity level and soil type.
Probabilistic Computing for Efficient Robotic Vision in Space17 Jan 2012 An integrated software and hardware approach for reducing the computational effort and energy expenditure of computer vision algorithms is developed. It is based on local sampling on the software side and probabilistic computing on the hardware side.
Evolution of adaptive behavior in a gravity-varying environment01 Nov 2011 Can we design a controller that can adapt to changing environmental conditions? If the environment is unknown in advance and dynamic, what is the best strategy to use for an agent and which methodology?
Scent of Science01 Oct 2011
Scent of Science01 Oct 2011 Can we use strategies similar to those of moths to find scientifically interesting locations in space? This study aims at investigating the use of similar strategies in connection to the localisation of different types of sources.
Evolutionary robotics for satellite swarms01 Aug 2011
Evolutionary robotics for satellite swarms01 Aug 2011 Is it possible to evolve the controllers of satellites in a swarm to bring about coherent swarm behavior? Does the evolutionary robotics approach have any advantages with respect to state-of-the-art control methods for satellites? In this project, the evolutionary robotics (ER) methodology was applied to the control of satellites in a swarm.
Time-To-Contact for spacecraft landing01 Jul 2011
Time-To-Contact for spacecraft landing01 Jul 2011 Can the time-to-contact be a useful visual cue to achieve robust and mass-efficient spacecraft landing?
On the Optimality, Adabtibility and Robustness of Neuro-controllers17 Apr 2010 Can we design a neuro-controller that can adapt to varying conditions? What is the tradeoff between the optimality of such neuro-controllers and their adaptibility and robustness?
Machine learning for global trajectory optimisation17 Nov 2009 Investigating the use of machine learning tools for the global optimisation of interplanetary trajectories.
Evolution in robotic islands01 Oct 2009
Evolution in robotic islands01 Oct 2009 Can we improve the automatic design methodology of Evolutionary Robotics by performing the optimisation of a neuro-controller in an island model?
WORKSHOP: Artificial Intelligence in Space (IJCAI-09)17 Jul 2009 The Advanced Concepts Team of ESA and the Artificial Intelligence Group of NASA's Jet Propulsion Lab organized a workshop on Artificial Intelligence in Space at IJCAI-09 in Pasadena, California, on July 17-18, 2009. The IJCAI is the largest AI conference in the world.
Clouds for modelling uncertainties in robust space system design01 Nov 2008 Traditional modelling of uncertainties faces several problems. The ACT is assessing a promising new approach for an autonomous and robust design, based on the concept of clouds combining the concept of a fuzzy set and that of a probability distribution.
Enhanced Situation Awareness01 Mar 2008
Enhanced Situation Awareness01 Mar 2008 How can we infer the health status of a spacecraft and the variation of its physical properties from the readings of few, strategically placed, sensors? The Advanced Concepts Team is assessing a novel approach to infer the health status of the spacecraft and its instrumentations from the readings of few, strategically placed thermal sensors.
WORKSHOP: Artificial Intelligence for Space Applications (IJCAI 07)08 Jan 2007 The Advanced Concepts Team, along with NASA's Jet Propulsion Lab and the Department of Computer Science of the University of Texas at El Paso, organized the workshop "Artificial Intelligence for Space Applications" at the International Joint Conference on Artificial Intelligence (IJCAI) in Hyderabad, India on the 8th January 2007.
Bayesian Inference for Asteroid Selection01 Jan 2007
Bayesian Inference for Asteroid Selection01 Jan 2007 Can we teach a machine to beat a human expert in solving the asteroid selection problem? The ACT is tackling the challange of employing Bayesian techniques for infering the optimal asteroid sequence.
Swarm Intelligence17 Apr 2006
Swarm Intelligence17 Apr 2006 Does a coherent group behaviour require an explicit mechanism of cooperation? Can useful tasks be accomplished by heterogeneous agents without direct communication and using decentralized control?
Natural Language Processing01 Dec 2004
Natural Language Processing01 Dec 2004 The aim of the study was to assess and produce an "intelligent advisor" tool for spacecraft designers.