A Neurorobotics and Spiking Neural Networks Approach to Cognitive Faculties of Reasoning and Planning

神经机器人和尖峰神经网络的推理和规划认知能力方法

• 批准号: 2270217
• 金额: --
• 依托单位: Heriot-Watt University
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2270217
• 关键词: Neurorobotics; Spiking; Neural; Networks; Approach

This work will explore the use of spiking neural network models based on observations of areas of the brain associated with cognitive capabilities. In particular reservoir computing models of the prefrontal cortex, and continuous attractor models of the hippocampus will be utilised. This contrasts with existing work which utilises a rule-based/symbolic model, or a non spiking artificial neural network. One key reason for not using the latter is to attempt to produce not only a functional model of cognitive capabilities, but a model which can potentially be used for studying the biological source. These models will be tested by being embodied in simulated and physical robotic agents (neurorobot) and evaluated on behavioural tasks.

这项工作将探索使用尖峰神经网络模型的基础上，与认知能力相关的大脑区域的观察。特别是水库计算模型的前额叶皮层，和连续吸引模型的海马将被利用。这与利用基于规则/符号模型或非尖峰人工神经网络的现有工作形成对比。不使用后者的一个关键原因是试图不仅产生认知能力的功能模型，而且产生一个可能用于研究生物来源的模型。这些模型将通过体现在模拟和物理机器人代理（神经机器人）进行测试，并在行为任务上进行评估。