BABEL

行为和计算神经科学，认知机器人技术和大规模神经网络的硬件实现的最新进展，提供了机会，加速了解大脑功能和设计基于大脑启发的控制系统的交互式机器人系统。在行动和语言学习领域，情况尤其如此，因为这一领域的科学和技术取得了重大发展。该项目旨在推进对单词学习中神经和行为机制的理解，在神经解剖学基础模型中验证这些原则，以及在SpiNNaker神经形态架构中实时实现大脑语言模型，以支持与神经成像实验的比较。科学假设和皮层语言模型也将通过在人形机器人iCub上实现体现主动语言学习的模型来验证。具体而言，在该项目中，我们将根据神经科学原理，在神经回路水平上开发语言学习理论，并构建语言皮层的神经计算模型，该模型实现了用于在大规模神经回路中学习用于谈论对象和动作的单词。这项理论工作将得到新的，假设驱动的脑成像研究的支持，使用MEG，EEG和功能磁共振成像，以确定神经相关和学习的对象和动作的单词的机制。成像结果将为大规模神经解剖模型的改进提供信息。这些模型将在SpiNNaker软件和硬件基础设施上实现，以使用更真实的尖峰活动来实现语言皮层的按比例放大的实时模型。最后，该项目将示范性地应用这些神经解剖模型和SpiNNaker系统作为人形机器人iCub的语言和动作学习模拟的控制器，在体现和主动学习环境中，语言的语义直接由对象操作任务的环境驱动。这是一个高度跨学科的项目，集成了神经形态工程的基本专业知识和方法，计算和实验神经科学，以及认知机器人。该项目是基于申请人的独特和战略伙伴关系，在这些专业领域的国际记录和以前的合作经验。此外，该项目还将受益于一个由学术和工业顾问组成的国际咨询委员会，以促进该项目的国际层面和影响。

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. This is especially the case in the domain of action and language learning, given the significant scientific and technological developments in this field. This 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 also be validated by implementing a model of embodied active language learning on the humanoid robot iCub. Specifically, in the project we will develop, based on neuroscientific principles, a theory of language learning at the neural circuit level and build a neurocomputational model of the language cortex that implements the learning of words used to speak about objects and actions in large-scale neuronal circuits. This theoretical work will be supported by novel, hypothesis-driven brain imaging investigations using MEG, EEG and fMRI to identify the neural correlates and mechanisms of the learning of words for objects and actions. Imaging results will inform the improvement of the large-scale neuroanatomical models. These models will be implemented on the SpiNNaker software and hardware infrastructure, to implement a scaled-up real-time model of the language cortex using more realistic spiking activity. Finally, the project will translationally apply these neuro-anatomical models and SpiNNaker system as controllers for language and action learning simulations with the humanoid robot iCub, within the embodied and active learning context where the semantics of the language is directly driven by the context of object manipulation tasks.This is a highly interdisciplinary project that integrates essential expertise and methodologies from neuromorphic engineering, computational and experimental neuroscience, and cognitive robotics. The project is based around the unique and strategic partnership of applicants with an international track record in these areas of expertise and with previous collaborative experience. Furthermore, the project will benefit from an International Advisory Board, with both academic and industrial advisors, to foster the international dimension and impact of the project.