Implicit and explicit communication for cooperation between humanoid robots

点击翻译按钮获取中文摘要

When cooperating on a task such as moving an object, humans will use a mixture of explicit communication (e.g. speech) and implicit communication (e.g. implied information through action) to ensure coordination. This research is investigating the relationship between implicit and explicit communication between robots engaged in collaborative tasks and aims to mimic in robots the intuitive way humans collaborate. By establishing implicit communication as a valid channel for information transfer, multi-robot systems could be fault-tolerant to environmental changes or hardware failure by conveying information through indirect channels. Systems could also become more efficient and economical by reducing the amount of information that must be transferred through conventional explicit means. The potential impact is on any multi-robot application where centralised coordination is unavailable or undesirable, including any application to which a robotic swarm or collective would be suited. There are three aims for this research that are each explored in part through the research methodology employed: firstly, to perform a meta-analysis of existing literature in related fields, secondly to replicate the techniques developed in existing literature within the context of this specific research question and finally the further development and novel integration of these techniques to create an entirely new behaviour inspired by human-human interaction. Three key areas of interest have been identified to achieve the desired behaviour. Firstly, developing a way for robots to communicate implicitly, potentially inspired by existing implementation of implicit communication within swarm robotics, including force consensus and imitation. Secondly, mimicking human intuition in similar collaborative tasks by implementing an internal simulator within the robot so that they can evaluate the consequences of their actions and make decisions for the best next action to execute. Finally, exploring incorporating the concept of dynamic leadership, dynamically assigning the role of leader to different agents depending on which is best suited to the role at any given point. This will facilitate a guided collaboration between two agents, and it is anticipated that the integration of these three areas will allow for a new form of communication and collaboration to arise. This research aligns itself with EPSRC's artificial intelligence systems, robotics and control engineering research areas.