Probabilistic analysis applied to robots

Abstract

Robots are increasingly being used in industry and starting their way to our homes as well. Nonetheless, the most frequently used techniques to analyze robots motion are based on simulations or statistical experiments made from filming robots’ movements. In this work we propose an alternative way of performing such analysis by using Probabilistic Model Checking with the language and tool PRISM. With PRISM we can perform simulations as well as check exhaustively whether a robot motion planning satisfies specific Probabilistic Temporal formulas. Therefore we can measure energy consumption, time to complete missions, etc., and all of these in terms of specific motion planning algorithms. As consequence we can also determine if an algorithm is superior to another in certain metrics. Furthermore, to ease the use of our work, we hide the PRISM syntax by proposing a more user-friendly DSL. As a consequence, we created a translator from the DSL to PRISM by implementing the translation rules and also, a preliminary investigation about its relative completeness by using the grammatical elements generation tool LGen. We illustrate those ideas with motion planning algorithms for home cleaning robots.