An experimental study of environmental complexity as seen by robots

Robots are being called upon to perform increasingly complex tasks in ever more complex environments. One popular strategy for designing robot control systems is to imitate creatures in the natural world. Since living organisms are usually adapted to the environment they exist in, the question becomes how can robots be designed to be adapted to their surroundings? To answer this, we must know how complex an environment is, or how well a robot can perform a given task in it. This paper examines a complexity measure that was previously proposed [36] for a robot searching task and shows the measure can predict the average number of steps a robot needs to complete a task to within 10%.