Abstract:

Robots are studied not only for their uses in laboratories and factories, but with an eye to incorporating them into human daily life. One of the commonest human social behaviors involves becoming a member of a queue, so for robots to integrate and co-exist with humans, queuing is an important skill to acquire. Given a camera and laser rangefinder, the robot will be able to map its environment, and it can then use Simultaneous Localization and Mapping, together with Moving Object Tracking (SLAMMOT), and a human detection algorithm to provide itself with a preliminary understanding of that environment.

Three categories of queuing model are proposed: with visible and invisible constraints, with partially visible constraints, and with no constraints within the environment. To decide the position to take relative to the rest of the queue, the algorithm uses curve fitting to perform the extrapolation, and takes into account invisible constraints such as personal space, as well as visible ones like obstacles and the direction of the queue. Where there are partial visible constraints, like stretch barriers, human behavior provides a cue to identify constraints that are hard for sensors to detect. The robot needs to observe what the humans are doing so that it can understand the constraints and behave in a way acceptable to the surrounding people. In an unconstrained environment, the robot can use the positions of human queue members as a reference, solving a linear programming problem to decide where to stand and the socially acceptable way to move toward it.