Cognitive control for decision and human-robot collaboration

This chapter focuses on the concept of cognitive control in robotics and how it is linked to decision, control and Human Robot Interaction (HRI). Achieving a control paradigm that enables robust, flexible goal-driven performance in a myriad of scenarios involving unstructured changing environments and interaction between robots and other agents such as humans has been sought during the last decade. In order to achieve this, inspiration has been taken from nature with a focus on the way humans and other animals undertake their decision and control processes. Indeed, by creating controllers inspired by human flexibility and adaptability, some or all the qualities found in human cognitive processes are pursued (i.e. adaptability, robustness, goal-driven behaviour with sensor and sub-task prioritization) in artificial programmable systems. Firstly, this chapter includes an introduction to the concept of control in the context of industrial processes, and expands it to robotics in general; challenges behind robot control will be raised, highlighting the need for novel decision and control architectures for modern robotics such as close human interaction, dealing with unstructured environments and learning to better perform a task - hence cognitive control. Secondly, the word “cognitive” in the context of control will be defined after an overview about how ”cognition” has been used in the literature beforehand; the definition of what a cognitive controller is will include aspects about both its architecture and inputs, highlighting how it relates to the term used originally in human behavioural studies and cognitive neuroscience. Finally, a modelling approach for cognitive control will be proposed, which integrates the principles of multi-agent interaction into a decision making (i.e. discrete and probabilistic) and control action (i.e. continuous and dynamic) framework. This is followed by a discussion around the framework’s elements and their wider impact in different areas of application such as autonomous driving, teleoperation and human-humanoid interaction