Modeling Teleoperated Robot Driving Performance as a Function of Environment Difficulty*

Modeling mobile robot driving performance is a challenging task, but a simple, accurate model can be helpful in making robot system design decisions that balance cost and performance. Driving performance can depend on a variety of factors including human ability, capabilities of the physical robot, and characteristics of the robot's environment (e.g. obstacles, terrain). This paper investigates teleoperated robot driving time for a task that includes obstacles in an environment. We hypothesize that a model analogous to Fitts' Law can be used to describe robot performance (driving time) as a function of environment difficulty. A new definition of the difficulty index (ID) in Fitts' Law is proposed in this paper that describes an environment's difficulty based on the arrangement of obstacles. The model is tested with data from a human subject study we conducted, in which a simulated differential drive robot was teleoperated through different environments in a manual control mode and semi-autonomous mode (obstacle avoidance). We demonstrate that the model for driving time and our proposed environment difficulty index fit human subject data well for a simple driving task between one pair of obstacles. Additionally, the model is expanded to predict driving time in a more complex environment with multiple pairs of obstacles. The results of this paper provide a building block for predicting teleoperated driving time performance in larger, more complex environments.