(18 Oct, 2010, Tapei, Taiwan)
Stefano Nolfi
Institute of Cognitive Sciences and Technologies, CNR
Via S. Martino della Battaglia, 44, 00185, Roma, Italy
This is the homepage of the tutorial on Evolutionary and Adaptive Robotics that will be held during the 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2010) at Taipei (Taiwan) on October 18, 2010.
The page describe the themes that will be illustrated during the tutorial and reference material.
The tutorial will be of interest for who want to know the method, the state-of-art in this field, and the advantages/drawbacks of different methodologies. The themes presented aim to be self-explanatory and do require previous knowledge. At the same time the tutorial aims to provide useful information also for those that are already familiar with evolutionary and/or developmental approaches to robotics and are interested in knowing the progresses achieved during the last years, the failures, and the challenges for the future.
With the term Adaptive Robotics we refer to evolutionary and/or developmental methods (Nolfi & Floreano, 2000, 2002; Weng et al., 2001, Lungarella et al. 2003, Oudeyer, Kaplan & Hafner, 2007; Asada et al. 2009) that allow to synthesize robots that evolve/develop their skills autonomously in interaction with the physical, and eventually social, environment on the basis of an adaptive process driven by the ecological condition in which the robot operate and by the feedback provided by the experimenter. This means that the strategy with which the robots solve their adaptive task as well as the detailed characteristics that allow the robots to display their skills in interaction with environment are shaped by the adaptive process (and are not programmed/designed by the experimenter).
Adaptive Robotics is particularly suitable for the development of robots that are embodied (i.e. that are able to exploit properties originating from the numerous interactions occurring over time between their body and the environment) and situated (i.e. that are able to exploit the properties arising from the fact that they can modify the next experienced sensory states through their action). In other words robots that, besides being provided with a physical body and beside being situated in a physical environment, are able to exploit the opportunities that they embodied and situated nature provides to them (Nolfi, 2009).Moreover, as we will demonstrate, Adaptive Robotics is particularly suitable for the development of embodied cognition skills, i.e. internal processing capabilities that are “grounded” on simpler behavioral and cognitive skills and, ultimately, on fine-grained sensory-motor interactions.
In the first part of the tutorial we will describe in more details which are the basic theoretical and methodological assumptions behind adaptive robotics approaches and which are the relation with other related approaches with particular reference to bio-inspired (Meyer & Guillot, 2008), behavior-based (Arkin, 1998), and other learning/developmentalapproaches. The methods that can be used to design adaptive robots will be described in more details in section 5.
In this section we will illustrate in which sense behavior and cognition in autonomous robots should be characterized as dynamical properties that emerge from a large number of fine-grained interactions occurring among and within the robot body, the robot’s control system, and the environment.
This will be illustrated theoretically and practically through three simple experiments involving: (i) passive walking robots (McGeer, 1990; Vaughan, Di Paolo & Harvey,2004), (ii) robots able to display simple navigation and discrimination skills (Nolfi, 2002), and (iii) robots able to self-localize in maze-like environments (Tani & Nolfi, 1999; Gigliotta & Nolfi, 2008). These examples will illustrate respectively: (i) how behavioral skills might emerge from the interaction between the robot body and the physical environment mediated only by the laws of physics, (ii) how simple behavioral/cognitive skills might emerge from the interactions between the robot and the environment mediated by simple reactive rules that regulate how the robot reacts to different sensory states, (iii) how simple cognitive skills might arise from the dynamical processes occurring within the robot’s neural controller and as a result of the robot/environment interactions.
For a discussion of the emergent nature of behavior see (Van Gelder, 1998; Beer, 2000; Keijzer, 2001; Nolfi, 2009) For additional information and video on passive walking robots see the following pages edited by Art Kuo,Steve Collins, and Eric Vaughan.
In this section we will illustrate in which sense behavior and cognition in autonomous robots should be characterized as dynamical properties with a multi-level and multi-scale organization in which: (a) the fine-grained interaction between the robot and the environment (including the social environment) give raise to low-level behaviors and in which the interaction between lower-level behaviors give rise to higher levels behaviors, and (b) higher level properties later affect lower level behaviors and lower-levels interactions.
This issue will be illustrated theoretically and practically through the description of experiments involving: (i) groups of robots that develop an ability to coordinate and cooperate, and communicate (Baldassarre et al, 2007), and (ii) robots able to develop and exhibit multiple behavioral skills as well as an ability to arbitrate them (Yamashita & Tani, 2008), and (iii) robots able to exploit the effects that the execution of a particular behavior has on the successive robot/environmental interactions (Nolfi & Marocco, 2002, Tuci, Massera & Nolfi, in press)
For more information on the multi-level and multi-scale nature of behavior and cognition see (Kelso, 2005; Yamashita & Tani, 2008; Nolfi, 2009). For more additional examples of the role of sensory-motor coordination, see (Scheier et al. 1998; Mirolli, Ferrauto & Nolfi, in press).
In this section we will try to demonstrate that adaptivity, defined as the ability of the robots to develop its own skills autonomously while it interact with the physical and social environment in which it is situated, represents a fundamental characteristics of autonomous robots. More specifically we will illustrate: (i) the relation between adaptivity, embodiment, and situatedness, (ii) the relation between adaptivity, the emergent nature of behavior and cognition, andthe multi-level and multi-scale organization of behavioral systems, (iii) the relation between adaptivity and “intelligence” with particular reference to the fact that one crucial aspect of intelligence is constituted by the ability to adapt to variations of the physical and social environment, (iv) the incremental nature of evolution and development with particular reference to the possibility to develop new abilities by re-using or by capitalizing on previously developed skills than thus act as scaffolds and to the possibility to enable a transfer of knowledge between different acquired capabilities.
Also in this case, this theme will be illustrated theoretically and practically through the description of experiments involving: (i) the emergence of communication in a population of evolving robots (de Greef & Nolfi, in press), (ii) the co-development of behavioral and linguistic skills (Sugita & Tani, 2005), (iii) the evolution of language in population of robots playing language games (Steels, 2003, 2010), (iv) the co-evolution or predator and prey robots (Floreano & Nolfi, 1997; Nolfi & Floreano, 1998).
In this concluding section we describe some of the most popular methods that are used to design adaptive robots, with particular reference to: evolutionary (Nolfi, and Floreano, 2000; Bongard & Pfeifer, 2003; Floreano, Husband & Nolfi 2008) and developmental (Weng, 2004; Oudeyer, Kaplan, & Hafner, 2007) and social learning methods (Steels, 2003, 2010). For each method we will describe the essential elements, the advantages, and the drawbacks.
We will concluding by discussion the challenges and the open problems in adaptive robotics research.
Arkin R.C. (1998). Behavior-Based Robotics.MIT Press, Bradford Book.
Asada M., Hosoda K., Kuniyoshi Y., Ishiguro H., Inui T., Yoshikawa Y., Ogino M., and Yoshida C. (2009). Cognitive developmental robotics: a survey. IEEE Transactions on Autonomous Mental Development, Vol.1, No.1, pp.12--34, 2009.
Baldassarre G., Trianni V., Bonani M., Mondada F., Dorigo M. & Nolfi S. (2007). Self-organised coordinated motion in groups of physically connected robots. IEEE Transactions on Systems, Man, and Cybernetics, 37(1):224-239
Beer, R.D. (2000). Dynamical approaches to cognitive science. Trends in Cognitive Sciences4(3):91-99.
Bongard, J. C. and R. Pfeifer (2003) Evolving Complete Agents Using Artificial Ontogeny, in Hara, F. and R. Pfeifer, (eds.), Morpho-functional Machines: The New Species (Designing Embodied Intelligence) Springer-Verlag, pp. 237-258.
De Greef J. & Nolfi S. (2010). Evolution of implicit and explicit communication in a group of mobile robots. In S. Nolfi & M. Mirolli (Eds.), Evolution of Communication and Language in Embodied Agents. Berlin: Springer Verlag.
Floreano D., Husband P. & Nolfi S. (2008). Evolutionary Robotics. In Siciliano B., Oussama Khatib (eds.), Handbook of Robotics, Berlin: Springer Verlag, pp. 1423-51
Floreano D. & Nolfi S. (1997). Adaptive behavior in competing co-evolving species. In P. Husband & I. Harvey (Eds), Proceedings of the Fourth Conference on Artificial Life, MIT Press, Cambridge, MA, 378-387
Floreano D., Mitri S., Magnenat S., Keller L. (2007) Evolutionary Conditions for the Emergence of Communication in Robots. Current Biology, 17 pp. 514-519
Gigliotta O., Nolfi S. (2008). On the Coupling Between Agent Internal and Agent/Environmental Dynamics: Development of Spatial Representations in Evolving Autonomous Robots, Adaptive Behavior, 16: 148-165
Keijzer F. (2001). Representation and behavior. London, UK: MIT Press
Kelso J.A.S. (1995). Dynamics patterns: the self-organization of brain and behaviour. CambridgeMA: MIT Press.
Lipson, H., Pollack J. B., (2000),Automatic Design and Manufacture of Artificial Lifeforms, Nature406:974-979
Lungarella M. Metta G., Pfeifer R., Sandini G., (2003). Developmental robotics: a survey. Connection Science,(15) 4: 1360-1494.
Massera G., A. Cangelosi & S. Nolfi (2007). Evolution of Prehension Ability in an Anthropomorphic Neurorobotic Arm, Frontiers in Neurobotics, 1(4):1-9.
McGeer, T. (1990). Passive walking with knees. In Proceedings of the IEEE Conference on Robotics and Automation, volume 2:1640–1645.
Meyer J-A., Guillot A. (2008). Biologically inspired robots. In Siciliano B., Oussama Khatib (eds.), Handbook of Robotics, Berlin: Springer Verlag, pp. 1395-1422.
Mirolli M., Ferrauto T., Nolfi S. (in press). Categorisation through evidence accumulation in an active vision system, Connection Science.
Nolfi S. (2002). Power and limits of reactive agents. Neurocomputing, 49:119-145.
Nolfi S. (2009). Behavior and cognition as a complex adaptive system: Insights from robotic experiments. In C Hooker (Ed.), Handbook of the Philosophy of Science. Volume 10: Philosophy of Complex Systems. General editors: Dov M. Gabbay, Paul Thagard and John Woods.Elsevier
Nolfi S. & Floreano D. (1998). Co-evolving predator and prey robots: Do 'arm races' arise in artificial evolution?Artificial Life, 4 (4), 311-335.
Nolfi S. & Floreano D. (2000). Evolutionary Robotics: The Biology, Intelligence, and Technology of Self-Organizing Machines. Cambridge, MA: MIT Press/Bradford Books.
Nolfi S. & Floreano D. (2002). Synthesis of autonomous robots through artificial evolution, Trends in Cognitive Sciences, (6) 1: 31-37.
Nolfi S. & Marocco D. (2002). Active perception: A sensorimotor account of object categorization. In From Animals to Animats 7. In B. Hallam, D. Floreano, J. Hallam, G. Hayes, J-A. Meyer (eds.) Proceedings of the VII International Conference on Simulation of Adaptive Behavior. Cambridge, MA: MIT Press, pp. 266-271.
Oudeyer P-Y, Kaplan , F. and Hafner, V. (2007). Intrinsic Motivation Systems for Autonomous Mental Development, IEEE Transactions on Evolutionary Computation, 11(2):265--286.
Scheier, C. et al. (1998) Embedded neural networks: exploiting constraints. Neural Networks, 11:1551-1596
Steels, L. (2003) Evolving grounded communication for robots. Trends in Cognitive Science. 7(7):308-312.
Steels L. (2010). Modeling the formation of language: Embodied experiments. In S. Nolfi & M. Mirolli (Eds.), Evolution of Communication and Language in Embodied Agents. Berlin: Springer Verlag.
Sugita Y. & Tani J (2005).Learning semantic combinatoriality from the interaction between linguistic and behavioral processes, Adaptive Behavior 13: 33-52.
Tani J. & Nolfi S., (1999). Learning to perceive the world as articulated: An approach for hierarchical learning in sensory-motor systems, Neural Networks, 12:1131-1141
Tuci E., Massera G., Nolfi S. (in press). Active categorical perception of object shapes in a simulated anthropomorphic robotic arm. IEEE Transactions on Evolutionary Computation,
Vaughan, E., Di Paolo, E.A. & Harvey, I., (2004). The evolution of control and adaptation in a 3D powered passive dynamic walker. In J. Pollack, M. Bedau, P. Husband, T. Ikegami & R. Watson (Eds). Proceedings of the Ninth International Conference on the Simulation and Synthesis of Living Systems. Cambridge, MA: MIT Press
Van Gelder, T. (1998). The dynamical hypothesis in cognitive science, Behavior and Brain Sciences 21: 615-665.
Yamashita Y. & Tani J. (2008). Emergence of functional hierarchy in a multiple timescale neural network model: a humanoid robot experiment, PLoS Computational Biology, 4 (11).
Weng J., McClelland J., Pentland A., Sporns O., Stockman I., Sur M. and Thelen E. (2001). Autonomous Mental Development by Robots and Animals, Science, vol. 291, 5504:599-600.
Weng J. (2004). Developmental Robotics: Theory and Experiments. International Journal of Humanoid Robotics, vol. 1, no. 2, 2004