An information pipeline model of human-robot interaction

This paper investigates the potential usefulness of viewing the system of human, robot, and environment as an “information pipeline” from environment to user and back again. Information theory provides tools for analyzing and maximizing the information rate of each stage of this pipeline, and could thus encompass several common HRI goals: “situational awareness” [6], which can be seen as maximizing the information content of the human's model of the situation; efficient robotic control, which can be seen as finding a good codebook and high throughput for the Human-Robot channel; and artificial intelligence, which can be assessed by how much it reduces the traffic on all four channels. Analysis of the information content of the four channels suggests that human to robot communication tends to be the bottleneck, suggesting the need for greater onboard intelligence and a command interface that can adapt to the situation.