| The debate over the value of mobile versus stationary code is an ongoing one. One possible way to quantify value is to determine the intelligence, that is, the ability to solve problems, of the different approaches. In the area of Artificial Intelligence in general, it is difficult to quantify and compare the quantity of intelligence of different methods as they perform a particular task. In particular, the relative intelligence, when solving the same task, of the mobile agent approach compared to the static agent (non-mobile) approach is difficult to quantify.;Other than the Turing Test, few metrics for measuring the intelligence of programs have been developed. We claim that intelligence without action is meaningless. As such, we are particularly interested in what happens at runtime. We define several measures of intelligence to be the inverse of runtime complexity for a given task. We use these metrics to examine the intelligence of groups of mobile and intelligent agents. Our examination is via case study using Whitebox Robotics PCBOT 914 robots. We implement a system with mobile agents and an equivalent system with non-mobile intelligent agents that both perform a typical robotic task. Two of the three metrics show much higher intelligence for the mobile agent system. The third metric is inconclusive. |
