Evolutionary approaches for improving the behaviour of robotic software components

改进机器人软件组件行为的进化方法

• 批准号: 2427590
• 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: 2020
• 资助国家: 英国
• 起止时间: 2020 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2427590
• 关键词: Evolutionary; approaches; improving; behaviour; robotic

This PhD project will investigate how methods from evolutionary computation could be used to improve both the functional and non-functional (e.g. efficiency, robustness, explainability) characteristics of existing robotic software components. A possible approach would involve using multi-objective evolutionary algorithms to generate improvements to an existing component's behaviour whilst simultaneously exploring the trade-offs within the component's design space. A potential focus will be on the improvement of deep neural networks embedded within the control and vision systems of robots.

这个博士项目将研究如何从进化计算的方法可以用来改善现有的机器人软件组件的功能和非功能（例如效率，鲁棒性，可解释性）的特点。一种可能的方法是使用多目标进化算法来改进现有组件的行为，同时探索组件设计空间内的权衡。一个潜在的重点将是改进嵌入机器人控制和视觉系统中的深度神经网络。