Supervisory control of multi-robot systems by disaggregation and scheduling of quasi-natural language commands

Natural language is a very intuitive and efficient way to interact with robotic systems and to control mobile multi-robot systems (MRS). Because of this we designed a graphical user interface (GUI) which allows the creation of orders as sentences. The interface allows only sentences in a quasi-natural language that follow production rules and keeps them unambiguous. Because of these properties those sentences can be easily processed by a computer. This concept allows an intuitive input of high level tasks while we avoid natural language processing problems like complex, ambiguous sentences. In this paper we show how those high level tasks can be mapped into basic robot commands. This mapping can require complex conditions, e.g., spatial or temporal constrains, and depends on the capabilities of the MRS. In this paper we focus on how the result of this mapping can be scheduled. We used genetic algorithms to determine which task should be executed by which robot and in what order. A planning algorithm calculates the expected start and end times of every task and tries to satisfy the given constrains.