Symbolic planning for industrial applications - the eRobotics approach

One of the greatest challenges in automated planning in industrial environments is to create the link between a real world scenario and symbolic representations that are used by a planner. In this paper we describe an approach to interface a planner with a close-to-reality simulation system. By using this approach, we move the problematic interface with a physical scenario to the simulation system, where we can benefit from a representation that is connected with the real world by methods being developed in the context of simulation-based control and eRobotics, an evolving branch of eSystems engineering. Along with these methods for grounding, we get full control over the system and its inner state, not only when the real system is running, but also during development in a purely simulated environment. Since the planner is completely decoupled from the environment, the connection to reality can be established when the development process is completed. We are using the Planning Domain Definition Language (PDDL) to interface different planners with our simulation system. Planning components in the context of industrial robotics were developed, which enable intuitive modeling of knowledge representations. A representation of actions using the custom Petri net scripting language SOML++ is used for intuitive visualization of generated planner output and for execution control in the physical and in simulated scenarios.