Intelligent Robotics

Brain-like Robot Learning

The idea of designing robots so that they learn through a developmental process, rather than by being programmed for a specific task, is now a fast growing new research area, known as Developmental Robotics [1]. We have several active research projects in this area and this PhD will form part of the group [2, 3].

The key idea is in getting robots to learn for themselves and we have built some experimental robot systems that gradually learn about their own sensory-motor systems and go on to discover how to coordinate and control themselves in order to learn more about their local environment. These robots progress through staged levels of learning and show a developmental life-cycle. During learning the robot may notice a stimulus that it may have seen before and will need to match this with its memory to retrieve further detail.

The goal of this project is to investigate mechanisms for sensory-motor memory that capture some of the characteristics of human behaviour, as given by current knowledge in psychology and brain science. This will be achieved by designing, implementing and experimenting with different algorithms for matching and storing patterns of sensory-motor activity. Other parts of the project include: reading relevant literature; using the experimental equipment, (this includes robotic hardware, special sensors, image processing systems, control and simulation software); on-line experiments with full learning robot system; analysis and evaluation; and report writing.

Proposed Supervisor: Prof Mark Lee. Other researchers in the group will also support this project.

References

[1] M. Lungarella, G. Metta, R. Pfeifer, and G. Sandini. Developmental Robotics: a Survey. Connection Science, 15(4):151-190, 2003.

[2] M.H. Lee & Q. Meng Psychologically Inspired Sensory-Motor Development in Early Robot Learning, International Journal of Advanced Robotic Systems, Volume 2, Number 4, 325-334, 2005.

[3] M.H. Lee and Q. Meng, Growth of Motor Coordination in Early Robot Learning, 19th International Joint Conference on Artificial Intelligence, IJCAI-05, Edinburgh, pp1732-33, 2005.

Brain Inspired models of Robot Development

This project is will be in Developmental Robotics [1, 2], which is a new and rapidly growing research field. Neuroscientists have produced a much greater understanding of the brain in recent years and computational models of brain systems are being built by many researchers worldwide [3]. In collaboration with colleagues in psychology, neuroscience and engineering at other universities, we are designing implementation models based on some of the sensory-motor structures in the human brain [2, 4]. These work on quite different principles from most computing concepts (for example, actions are performed not by activating a motor system but by stopping the inhibition of a motor system!) and it seems likely that these new control methods will turn out to be valuable for robotics and computing in general.

The goal of this project is to investigate certain mechanisms for sensory-motor learning that capture some of the characteristics of human behaviour, as given by current knowledge in psychology and brain science. This will be achieved by designing, implementing and experimenting with different sensory-motor learning algorithms. It seems that sensory-motor learning comes before all other learning [1] and so such a fundamental process could be very useful for robot learning. There will be various stages to the work: background reading; using the experimental equipment, (this includes robotic hardware, image processing systems, control and simulation software); design and implementation of algorithms; on-line experiments with full learning robot system; analysis and evaluation; and report writing.

Proposed Supervisor: Prof Mark Lee. Other researchers in the group will also support this project.

References

[1] R. Pfeifer and C. Scheier, Sensory-motor coordination: the metaphor and beyond, Robotics and Autonomous Systems, 20(2):157-178, 1997.

[2] M. H. Lee and Q. Meng, Growth of Motor Coordination in Early Robot Learning, 19th International Joint Conference on Artificial Intelligence, IJCAI-05, Edinburgh, pp1732-33, 2005.

[3] A. Clark. Being There - Putting Brain, Body, and World Together Again, MIT Press, Cambridge, MA, 1998.

[4] Q. Meng, and M. H. Lee, Novelty and Habituation: the Driving Forces in Early Stage Learning for Developmental Robotics, in Biomimetic Neural learning for intelligent robotics, LNCS. Vol 3575, pp315-33, 2005.

Autonomous Robot Planetary Scientist

The Mars Exploration Rovers (MER) [Squyres 04] have achieved a 10 fold increase in traverse distance travelled per Sol when compared with the first rover to land on Mars [Stone 96], and are now in a further extended mission phase. This success is not without cost as primary mission operations costs can be often as much as 40% of the total primary mission cost. The fundamental problem is that large teams of scientists and engineers are involved in the tasks of defining, rehearsing, planning, scheduling and uploading every single activity associated with surface operations, no matter how small or large that activity might be. The ESA ExoMars rover is planned to traverse further and for longer than the MER rovers, and therefore these operations problems will arise. If they are not solved, then potential science data will be lost and operations costs will soar.

The ability for a rover to operate autonomously is advantageous as this could potentially increase science data return whilst reducing operations costs [Rajan 00]. Current work in the area of autonomous science data gathering for Earth observation satellite operations is showing how successful an autonomous approach can be [Chien 05]. However to apply such an approach to planetary surface rovers is not trivial, and JPL are beginning to invest in this area [Estlin 05]. Even to just maintain science data return parity means that Europe must also invest in this technology. Within Europe autonomous science technology for rover operations is an area in which the UK could take the lead.

The space and planetary robotics group has recently attracted £250,000 from the UK Higher Education Funding Council for Wales (SRIF3) to create a new 'Mars Yard'. Refurbishment is complete and a new indoor Mars analogue terrain facility has been created. This has resulted in a 100 m2 facility (45 m2 of terrain) which supports advanced rover and manipulator motion tracking equipment, and the latest operator/robot interaction and visualisation equipment. The Mars Yard will contain a new 'Concept E' rover chassis with a mounted manipulator, and associated sensors and instruments (e.g. cameras).

Supervisor: Professor David P. Barnes

References

[Squyres 04] Squyres S.W., et al, The Spirit Rover's Athena Science Investigation at Gusev Crater, Mars, Science, 305: pp. 794-799, August 2004. [Stone 96] Stone H.W., Mars Pathfinder Microrover a Small, Low-cost, Low-power Spacecraft, Jet Propulsion Laboratory, 1996. [Rajan 00] Rajan K. and Shirley M. and Taylor W. and Kanefsky B, Ground Tools for the 21st Century, Proceedings of IEEE Aerospace Conference, Big Sky, MT, 2000. [Chien 05] S. Chien, R. Sherwood, D. Tran, B. Cichy, G. Rabideau, R. Castano, A. Davies, D. Mandl, S. Frye, B. Trout, S. Shulman, D. Boyer, Using Autonomy Flight Software to Improve Science Return on Earth Observing One, Journal of Aerospace Computing, Information, and Communication, April 2005. [Estlin 05] Estlin T., Judd M., Gaines D., Castano A., Bornstein B., Stough T., Wagstaff K., Anderson R.C., Opportunistic Science with a Rover Traverse Science Data Analysis System, Proceedings 8th International Symposium on Artificial Intelligence, Robotics and Automation in Space (i-SAIRAS), Munich 2005. CD-ROM Proceedings.

Advanced Image Processing and Visualisation Methods for the ESA ExoMars Rover

The ExoMars rover [Barnes 06a] is the first element of the ESA Aurora programme and is scheduled to deliver the Pasteur exobiology payload to Mars by 2015. A key instrument on the rover is the 0.7 kg Panoramic Camera (PanCam) [Griffiths 06]. The PanCam imaging experiment is designed to obtain high-resolution colour and wide-angle multi-spectral stereoscopic panoramic images from the mast of the ExoMars rover (as part of the Pasteur payload). The PanCam has been designed to fulfil the digital terrain mapping (digital elevation models - DEM) requirements of the mission. For the proposed ExoMars PanCam instrument, it is essential that software tools are available to scientists (and engineers) to model, simulate and visualise returned camera data within the context of the ExoMars rover and surrounding Mars terrain. A number of basic software tools are required including: DEM range maps; surface normal data; instrument deployment reachability maps; surface roughness maps; DEM slope products; solar energy maps; terrain mesh products; image mosaic products, and terrain visualisation projections. Whilst the JPL Multimission Image Processing Laboratory (MIPL) has created similar software products for the NASA twin Mars Exploration Rover (MER) mission [Alexander 06], it is not the intension of this project simply to duplicate the JPL products. On the contrary, many of the above software products have already been developed by Aberystwyth University as part of our Beagle 2 work [Barnes 03]; other research council funded activities, ESA [Barnes 06b] and industry funded work. This project seeks to build upon our developed software and provide new science enabling capabilities for the ExoMars mission planetary scientists. We intend to include advanced computer graphics methods within the proposed modelling and simulation software tools. We propose to research and implement visualisation tools that will allow both rapid science data comprehension, and science target identification. Supervisor: Professor David P. Barnes

References

[Barnes 06a] Barnes D., et al., "The ExoMars rover and Pasteur payload Phase A study: an approach to experimental astrobiology", International Journal of Astrobiology. doi:10.1017/S1473550406003090, 2006.

[Griffiths 06] Griffiths A.D., et al., "Context for the ESA ExoMars rover: the Panoramic Camera (PanCam) instrument", International Journal of Astrobiology, doi:10.1017/S1473550406003387, 2006.

[Alexannder 06] Alexander D.A., et al., "Processing of Mars Exploration Rover imagery for science operations planning", Journal of Geophysical Research Vol. III, E02S02, doi:10.1029/2005JE002462, 2006.

[Barnes 03] Barnes D.P., Phillips N., and Paar G., "Beagle 2 simulation and calibration for ground segment operations". In Proc. 7th Int. Symposium on Artificial Intelligence, Robotics and Automation in Space, iSAIRAS'03, NARA Japan, May 2003, CD-ROM Proceedings.

[Barnes 06b] Barnes D.P, et al, "Imaging and localisation software demonstrator for planetary aerobots", Acta Astronautica, 59, pp. 1062-1070. Elsevier. doi:10.1016/j.actaastro.2005.07.050, 2006.