PI and Project Coordinator; Cognitive robotics
Recent advances in behavioural and computational neuroscience, in cognitive robotics, and in the hardware implementation of large-scale neural networks, provide the opportunity for an accelerated understanding of brain functions and for the design of interactive robotic systems based on brain-inspired control systems. The BABEL project aims at advancing the understanding of neural and behavioural mechanisms in word learning, the validation of these principles in neuroanatomically grounded models, and real-time implementations of brain language models within the SpiNNaker neuromorphic architecture that will support comparisons with neuroimaging experiments. The scientific hypotheses and cortical language model will be validated by implementing a model of embodied active language learning on the humanoid robot iCub.
PI and Project Coordinator; Cognitive robotics
Principal investigator; Neuromorphic engineering
Plymouth University and
Freie Universitate Berlin
Brain imaging, computational neuroscience
Spiking neural networks
Plymouth University and Freie Universitate Berlin
Computational Neuroscience Modelling;
The final project workshop will be in Manchester, on 7 October 2016. The BABEL Workshop on large-scale brain models for embodied language on robots using SpiNNaker will see talks from the project partners, as well as from invited speakers. Call for Participation here
The robotic experiments will be based on the iCub humanoid robot platform available at the Plymouth lab. The iCub is an open source platform for developmental robotics (Metta et al. 2010), based on a child-like morphology, with 53 degrees of freedom. The iCub will be used for language comprehension experiments, using the speech recognition software and artificial vision processing routines previously used by the Plymouth team (Morse et al. 2010; Tikhanoff et al. 2011). These tools are currently integrated in the Aquila software, linked to the YARP software interface for the robot hardware.
The neuromorphic SpiNNaker system employs a lightweight address-event representation (AER) packet-switched communication system to allow up to a million ARM processors to cooperate in real-time implementations of point-neuron models. 2-processor SpiNNaker test chips were fabricated in 2009 and support models of up to a few thousand neurons. An advanced 18-processor SpiNNaker chip was developed in 2011. This machine development is supported under the EPSRC BIMPA grant EP/G015740/1. These machines provide the hardware platform for the work proposed here. In addition, BIMPA will deliver software systems supporting various neuron (e.g. Izhikevich model) and synapse (STDP, NMDA) models, mapping large-scale networks (using PyNN) onto the machine’s resources, thus providing a generic platform for real-time neural models.
Media files will be made available here during the project.
Open source software will be added to the icub.org repository and linked here.
The project staff will collaborate with the industry advisors to support the impact and knowledge transfer of the BABEL research results and technologies to R&D areas in robotics, automation and microeletronics.
The companies and institutions directly involved in the BABEL impact strategy, through the role of indutry advisors of their senior staff, are HONDA, TSMUK, ARM and Cogniscience. In addition, the collaboration with the IIT will focus on open source developments of the BABEL results.
HONDA RESEARCH INSTITUTE EUROPE (Advisor Edgar Koerner, President of HRI EU). This is one of three fundamental research institutes of Honda worldwide. In Europe, the Instute aims for a better understanding and successful application of intelligent systems in an engineering and a business context, with specific focus in (i) Brain-like Intelligence; (ii) Evolutionary and Learning Technology; (iii) Embodied Brain-like Intelligence. In BABEL the collaboration with HONDA will focus on the application of the project results and technology in the car industry and humanoid and mobility robotics.
TMSUK Japan (Advisors Yoichi Tamamoto, Chiefe Executive and Founder, and Junji Matsuo, European R&D Officer). This is one of the leading international companies in service robotics. TMSUK was established in 2000, in Fukuoka, Japan, and since then it has developed over 30 robot platforms for search and rescue, telecomms, medical/dentist training and care companions.Such platforms are sold to government agencies, private companies and reserach institutions. In BABEL the collaboration with TMSUK will focus on the application of the project results and technology in the service robotics industry.
ARM Ltd (Advisor Ian Phillips, Principal Staff Engineer). ARM Ltd. This is a UK company based in Cambridge, with ~£400m revenue and 2,000 employees world-wide. Known as the world’s leading supply of System-on-Chip Intellectual Property (SOC-IP), its 32-bit RISC CPU family powers more than 95% of the smart electronic products which pervade our lives today. In BABEL the collaboration with ARM will focus on the application of the project results and technology in microelectronics and neuromorphic engineering.
Cogniscience Ltd (Steve Furber, Founder). This is a University of Manchester spin-out company established to commercialize the IP arising from the SpiNNaker project. This company is positioned to supply SpiNNaker hardware to support any exploitation of the results of the BABEL project that require the use of SpiNNaker hardware.
IIT (Advisor Giorgio Metta, Director of iCub Facility). The Italian Institute of Technology aims to promote Italy's technological development and advanced education, consistent with national policies for scientific and technological development, with particular emphasis on robotics, neuroscience, biomedical technology and nanotechnology. In BABEL the collaboration with the IIT iCub Facility and the Department of Robotics, Brain and Cognitive Sciences will focus on the application of the project results and technology in open source for humanoid robotics.