We held a successful workshop at IROS on the topic of Shared Autonomy. It brought together different perspectives and approaches on how to deal with (multiple) autonomous agents and how to organize their behavior.
As a further step, details and simulation about the action selection network have been presented at AISB 2017:
Schilling, M. (2017). “Old Actions in Novel Contexts — a Cognitive Architecture for Safe Explorative Action Selection”. Proceedings of the Artificial Intelligence and Simulation of Behaviour Conference (AISB 2017), Bath (UK).
Finally, the paper on our minimal cognitive system appeared: it describes the underlying embodied and biologically inspired system for a walking robot. And it introduces the cognitive expansion part: at first, how the internal model can be recruited for internal simulation. And second, how the overall processing structure is organized.
Schilling, M. and Cruse, H. (2017), “ReaCog, a Minimal Cognitive Controller Based on Recruitment of Reactive Systems”. Front. Neurorobot. 11(3). doi: 10.3389/fnbot.2017.00003
Recently, my focus turned more and more towards interaction. This leads to the question how different autonomous systems can, first of all, coexist in a shared space and deal with shared resources. In the long run, the goal is of course to extend this towards joint and collaborative behavior. But while this requires implicit and explicit coordination, it also requires that agents align their autonomous behavior and respect constraints given through other agents.
Schilling, M., Kopp, S., Wachsmuth, S., Wrede, B., Ritter, H., Brox, T., Nebel, B., Burgard, W. (2016). “Towards A Multidimensional Perspective on Shared Autonomy”. Proceedings of the AAAI Fall Symposium Series 2016, Stanford (USA).
We presented our work on a framework for the humanoid robot Nao at ICDL Epirob 2016 in Paris. The framework allows to setup game like interactions with the robot and was used for a small series of experiments. During these experiments, pre-schoolers interacted with Nao in a learning task that was designed as a small game.
Lücking, P., Rohlfing, K., Wrede, B., and Schilling, M. (2016). “Preschoolers’ Engagement in Social Interaction with an Autonomous Robotic System”. Proceedings of the Joint IEEE International Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob) 2016, Paris.
A perspective on the importance of mental simulation for minimal system approaches has appeared in PNAS:
Schilling, M. and Cruse, H. (2016), “Avoid the hard problem: Employment of mental simulation for prediction is already a crucial step”. Proceedings of the National Academy of Sciences (PNAS) 2016 ; published ahead of print June 28, 2016, doi:10.1073/pnas.1607146113.
We published a joint article on how representations are enriched during development. The article offers a detailed review on developmental work with children and how their representations develop, especially before language.
Nomikou, I., Schilling, M., Heller, V. and Rohlfing, K. J. (2016), “Action and interaction as contexts for enriching representations”. Interaction Studies, 17(1), pp. 128–153
Together with our chapter on cognitive properties of the minimal cognitive system, a commentary appeared on the emergent properties of such systems. It is part of the Open Mind (online) publication series:
Cruse, H. and Schilling, M. (2016), “Mental states as emergent properties. From walking to consciousness”. In: Open Mind, Philosophy and the Mind Sciences in the 21st Century. Vol 1. Metzinger T, Windt JM (Eds); Cambridge, Mass.: The MIT Press: 349-386.
There was a nice workshop on “Advances in Biologically Inspired Brain-Like Cognition and Control for Learning Robots” at IROS. I presented the work on recruiting the internal model for planning ahead and cognition as an extended abstract and poster there. The abstract can be found here.
We applied the MMC approach to a population-code based representation – this allows to decompose complex robotic structures into simple, local geometric relations that can be easily solved in parallel. The population-code representation shall be utilized in the future to allow for integration of additional sensory signals. The paper has been presented at the IJCNN in Ireland, 2015:
Population based encodings allow to represent probabilistic and fuzzy state estimates. Such a representation will be introduced and applied for the case of a redundant manipulator. Following the Mean of Multiple Computations principle, a neural network model (PbMMC) is presented in which the overall complexity is divided into multiple local relationships. This allows to solve inverse, forward and mixed kinematic problems. The local transformations in between the kinematic variables can be sufficiently well learned by small single MLP layers. … the model as such is quite flexible as it can keep track of multiple possible solutions at the same time.
Citation: Baum, M., Meier, M., and Schilling, M. (2015), “Population based Mean of Multiple Computations Networks: A Building Block for Kinematic Models”. International Joint Conference on Neural Networks 2015, Killarney (Ireland).