Research Topics
[MA] Automatic Task Assignment in Colored Petri-Net-based Dynamic Robot Swarm Collaborations
Robots are a promising technology to support humans various tasks. In industry robots are established for a long time (e.g. assembly) and can execute recurring tasks fast and reliable. Another field of application is the support of handicapped people in their daily life. In contrast to traditional industrial robots, software for service robots must be highly adaptive w.r.t. the structure and behavior of the software. They must operate in environments which are not known at design time and must fulfil tasks that cannot completely be specified at design time. Current software development approaches lack of concepts to design and implement software for robots that will reliably operate in any environment, for every user with individual requirements and in addition are able to model dynamic variation and extension of the provided functionality. A second problem is the diversity of the robotic hardware platforms. Current robots are not built physically be able to fulfil any tasks. In contrast specialized robots are built, that are able to execute a small set of services (e.g. cleaning, transportation) very reliably. In order to provide more complex services, the abilities of several robots have to be combined. One approach to model such ad-hoc collaborations is SwarmControl, developed by Till Kolditz <1>. Kolditz showed, that Hierarchical Colored Petri Nets can be used to model the execution of tasks severed by one robot and the overall behavior of several robots by composition of petri nets. A formal knowledge representation allows to model tasks semantically and express different levels of abstraction. In his work, a task can further be divided into subtasks. The execution of a task involves the assignment of the subtasks to one or more robot agents. Currently it is only possible to define services of a robot that are matched witch subtasks by String comparison. In order to decide whether or not a robot is able to fulfil a subtask, a detailed description of the capabilities of the robot (e.g. the ability to grab objects with dimension and weight) as well as a detailed description of the specific task (e.g. dimensions and weight of the object that should be grabbed) has to be provided. Furthermore context-dependent information has to be taken into account (e.g. current location of the robot), in order to decide which robots must be involved in the task solving process.
Betreuer: Christian Piechnick