ARS: First of all we would like to thank you for accepting to make the interview. Could you please briefly describe your scientific career and in that way present yourself.
Ijspeert: I originally studied physics, and received my diploma (equivalent to a BSc/MSc) from the EPFL, the Swiss Federal Institute of Technology, in Lausanne, Switzerland. I then decided to do research in neural computation and evolutionary algorithms in the Department of Artificial Intelligence at the University of Edinburgh, where I got my PhD. After that, I carried out two post-docs, one in collaborative robotics at the EPFL (in collaboration with IDSIA), and one in neural computation at the University of Southern California (USC), in Los Angeles. In 2000, I became a research assistant professor at USC. During that time, I also started collaboration with ATR ( Kyoto, Japan) in humanoid robotics. In 2002 I moved to my current position as a Swiss National Science Foundation assistant professor at the EPFL, where I am heading the Biologically Inspired Robotics Group (BIRG).
ARS: Tell us something about your current research and about your personal goals?
Ijspeert: I am interested in biologically inspired robotics. My goals are to use robots to gain a better understanding of biological control principles, and to use inspiration from biology to design robots that replicate animals’ agility and autonomy. In particular, I am fascinated by the agility exhibited by animals in controlling their movements and their locomotion. To me, this is an area in which robotics has lots to gain from taking inspiration from the animal kingdom. But note that I do not advocate blindly copying animal structures and control mechanisms (since they are the result of a “messy” evolutionary process and might not always be optimal), only those features that can make robots better. Research currently carried out in my group is in five areas: neural modelling of animal locomotion control, dynamical systems, amphibious robotics, humanoid robotics, and modular robotics. The neural modelling projects aim at understanding the neural control mechanisms of vertebrate animals, in particular central pattern generators, using numerical simulation tools. The dynamical systems projects explore how nonlinear dynamical systems (e.g. nonlinear oscillators) can be used for control, and made adaptive using insights from synchronization properties. The goal is to develop dynamical systems that are capable of learning, and for which the learning process is embedded into the dynamical systems framework (i.e. learning is part of the controller, not an external algorithm). The amphibious robotics project aims at designing amphibious snake (or lamprey) robots that are capable of switching from swimming to crawling. The humanoid robotics project explores how dynamical systems can be used for learning and controlling movements in a learning-by-demonstration framework. Finally, we recently started some efforts in modular robotics (i.e. robots that are composed of multiple robotic units than can attach and detach). The challenge here is to design efficient distributed controllers that can deal with a robot structure that changes over time.
ARS: How did you become connected with robotics?
Ijspeert: I started working in robotics during my PhD studies. My first experiments in robotics were the construction and control of Lego robots for a robot-rugby competition held at the department of Artificial Intelligence in Edinburgh. But otherwise my interest in understanding the mechanisms underlying animal movement and locomotion control naturally led me to robotics. While my PhD thesis was done in simulation, it involved designing neuromechanical simulations (i.e. simulation of both the neural structures and the biomechanics) of lamprey and salamander. The mechanical simulations were articulated rigid bodies models very similar to those one would make of a robot. One of our current projects, the construction of an amphibious snake robot, is therefore a direct follow up of those experiments in simulation. From there on, I got interested in the adaptive control of any type of robot with multiple degrees of freedom (e.g. quadruped, humanoid, and modular robots).
ARS: Since you have a significant international experience we would like to hear your impressions about robotic research in Europe, USA and Japan. What is the main research difference between these three continents?
Ijspeert: I was indeed fortunate to work in these three different environments. But I do not feel that I know them well enough to properly answer your question. I will however try to give an approximate answer. I think a rough, probably caricatured, picture is that robotics research in Japan is driven towards helping society (e.g. the elderly people), research in the USA is driven towards defence and military applications, and research in Europe is driven towards industry and science. These different orientations can be found in the funding agencies. In the USA, for instance, a large majority of projects are funded by DARPA or ONR (i.e. defence related agencies), and only a very minor part by the NSF. It is the opposite in Europe where funding comes mainly from national science foundations and from the European Union. The European Union often favours projects that involve some technology transfer to the industry, showing its dedication to reinforcing links to the industry.
ARS: Why is USA so in front comparing it with Europe? Is it in way of organisation, different approach in research or just difference in budget amount?
Ijspeert: I think that the USA are less in front than it might appear at a first look. The perceived difference often comes from the fact that American institutes are much stronger at “selling” and “marketing” their results than European institutes. But the US research environment has many very positive aspects such as dynamism (young researchers can have stellar careers if they have good ideas and work hard), originality (large freedom to try new research topics, especially interdisciplinary research), and good funding. And there is a kind of positive feedback effect in the fact that many bright researchers are attracted to work in the US because numerous others already work there.
ARS: You also have experience in working with Japanese robotic researchers. There is a feeling that they are a little bit closed to the world community. I mean, there are not so much scientific books written by Japanese authors published in English or some other non-Japanese language, but there are so many Japanese robotic researchers that are world-class scientists and surely have a lot of results to publish.
Ijspeert: I do not agree with you that Japanese robotics researchers are so closed to the world community. I think that language is a big handicap, and that western researchers would have the same difficulties if they had to publish and present their results in Japanese. And it seems to me that things are changing fast, with new generations of PhD students speaking a more and more fluent English. There are also some research domains, such as humanoid robotics, where a large majority of researchers are Japanese, and in those domains, it might be that the motivation to publish in English is lower than for others.
ARS: Back to Europe, is there a significant brain drain in European robotics community right now? In which direction by your opinion European robotic research community should go? Which advantages Europe has and USA and Japan not?
Ijspeert: These are difficult questions to answer. Concerning the brain drain, there is indeed a strong motivation to go abroad (mainly to the USA) for gaining international research experience. In Switzerland, for instance, it is almost impossible to get a faculty position without having spent some time in the USA. And since the USA offer such a dynamic environment, many researchers quickly obtain interesting positions, with the risk of not returning. But among colleagues and friends, I know many people who are returning to Europe, or are willing to come back, as soon as a good opportunity offers itself. I think this brain “gain” is strong enough to largely compensate the brain drain. To me, the best way of promoting European research is to offer competitive faculty positions to young researchers in robotics (e.g. by increasing the number of tenure-track assistant professorships). Europe will always be a nice place to do research because of its world-class academic institutes combined with excellent standards of living. And the European Union has launched several interesting funding initiatives (e.g. the “Beyond robotics” and “Cognitive systems” programs), which means that it is currently a good place for robotics research.
School of Computer and Communication Sciences,
EPFL Swiss Federal Institute of Technology Lausanne , Switzerland
Published in: Volume 2, Number 2, June 2005
For further informations in German please click here!