Petri Net Robotic Task Plan Representation: Modelling, Analysis and Execution

Abstract

Petri nets provide a practical and intuitive way of modelling robotic tasks and associated components, being also appropriate to monitor the execution of tasks given their graphical nature. The fact that a GSPN is equivalent to a Markov chain brings an additional advantage by allowing the use of currently available tools and techniques to extract important a priori information about a given task. Being able to model the actions more thoroughly at a lower level allows for mores realistic models, without compromising the analysis possibilities. Furthermore we can create all the models separately and build the task plan by creating a network of actions. This task plan can be ran directly on the robots for execution purposes and, for analysis purposes, we compose all the models that were designed separately onto one single Petri net, and analyse that net. The introduction of communication models allowed the extension of the framework to multirobot tasks, enabling a priori extraction of qualitative and quantitative properties of multirobot tasks. Different communication models enable the study of the impact of a range of communication problems on the task success. We are currently improving the communication