Algorithmic navigation to train deictic mobile robot operators

Presents an algorithm for a simulator to train mobile robot operators in the use of deictic commands. It shows that robots can be deictically controlled in any reasonable environment. Deictic control is a technique which directs the robot by visually servoing on targets pointed out to the robot by its operator. This provides a direct, intuitive control method which any sighted operator can use. Previous work has formally defined canonical targets required for deictic control, and has described a training system for prospective users. This paper extends that work by providing a specific algorithm for the construction of deictically controlled paths. The algorithm is described in detail and results are presented. This algorithm shows that deictic commands and their targets can be simply determined by an agent able to sense the surrounding environment