A Data-driven Approach Towards Human-robot Collaborative Problem Solving in a Shared Space
This work addresses human-robot interaction challenges in shared environments, but it is incremental as it focuses on data collection and simulation stages without final robot deployment.
The paper tackles the problem of enabling natural human-robot communication for collaborative problem-solving in shared spaces by developing a data-driven system, with initial results including the creation of data collection instruments and performance metrics for evaluation.
We are developing a system for human-robot communication that enables people to communicate with robots in a natural way and is focused on solving problems in a shared space. Our strategy for developing this system is fundamentally data-driven: we use data from multiple input sources and train key components with various machine learning techniques. We developed a web application that is collecting data on how two humans communicate to accomplish a task, as well as a mobile laboratory that is instrumented to collect data on how two humans communicate to accomplish a task in a physically shared space. The data from these systems will be used to train and fine-tune the second stage of our system, in which the robot will be simulated through software. A physical robot will be used in the final stage of our project. We describe these instruments, a test-suite and performance metrics designed to evaluate and automate the data gathering process as well as evaluate an initial data set.