Robotics for carbon-neutral fusion-energy - Understanding, modelling, and predicting the human tele-robotics operator for intelligent shared control

这个项目将通过身体运动和生理数据分析来调查遥控机器人系统操作员对动态决策和行动的感知的使用。通过这种分析得到的模型将被用来理解和预测操作员，目的是隔离操作员故障，并改善操作员和远程操作系统之间的互信。这也可能导致操作员通过智能共享控制方案的支持。这个问题的主要用例是英国原子能管理局(UKAEA)的吉祥物遥操作系统。吉祥物是一种双臂主从机器人，经过数千小时的成功远程操作，在JET(联合欧洲环)聚变反应堆的重新配置和维修任务中得到了验证。我们提出了一个使用身体跟踪设备对吉祥物操作员进行任务预测的系统。将采用机器学习方法进行适当的决策和行动建模(例如使用马尔科夫模型)，以便对操作员未来的任务进行分类和预测。在第二步，这个项目可能会研究触觉反馈的适应性。相信这种适当和任务特定的反馈可以促进吉祥物操作者在远程处理任务时的安全和效率，并通过积极引入增强触觉效果和其他动作来减少人为错误，从而提高吉祥物的安全和效率。这一博士学位将有助于对人-机器人相互作用的一般理解，介绍关于人类在这种过程中的决策和行动模型的新结果，并可能有助于共享控制和半自主的原则。

This project will investigate the use of the perception of dynamic decisions and actions of operators of telerobotic systems through body motion and physiological data analysis. The models obtained through this analysis will be used to understand and predict the operator with the aim to isolate operator failure and improve mutual trust between the operator and the tele-operation system. This may also result in the support of the operator through intelligent, shared control schemes. The major use case for this problem is the MASCOT tele-manipulator system of the UK Atomic Energy Authority (UKAEA). MASCOT is a two-armed master-slave robot established and proven through thousands of successful hours of remote handling operations in reconfiguration and repair tasks within the JET (Joint European Torus) fusion reactor. We propose a system for task prediction of the MASCOT Operators using body tracking devices. A machine learning approach for suitable decision and action modelling (e.g. using Markov models) will be deployed to classify and predict future tasks of the operator. In a second step, this project may look into the adaptation of haptic feedback. It is believed that this appropriate and task-specific feedback may promote safety and efficiency of MASCOT operators during remote handling tasks and reduce human errors by actively introducing augmented haptic effects and other actions for safety and efficiency of MASCOT. This PhD will contribute to the general understanding of human-robot interaction, introduce new results on human decision and action models in such processes and may contribute to principles of shared control and semi-autonomy.