Environment representation in the form of an ontology and object maps

In the proposed representation, it is supposed that each object is of some predefined type. Object is determined by a collection of attributes, and optionally its internal hierarchical structure consisting of sub-objects and relations between these sub-objects. The objects that do not have internal structure are called elementary objects, and their types are called elementary types. Object of type WALL may serve as an example of elementary object. It is determined by the attributes: width, height and color. The type ROOM may be example of a complex type; its internal structure is composed of elementary types such as walls, floor, ceiling, windows and doors, as well as the relations between these objects. A particular object (as an instance of a type) is defined by specifying concrete values of its attributes, and (if it is of complex type) also by specifying its sub-objects. Instance of the representation (called also model or a map of the environment) is defined as a specification of an object of a complex type.

An ontology for indoor environments was defined. It consists of types of objects commonly occurring in that kind of environments. The defined ontology is used by algorithms for map updating and its refinement as a priori knowledge. Based on the definitions of types, restrictions specified for possible values of their attributes and data form perception module, objects in the device surrounding can be classified or recognized. The classified objects can be added to an existing object map of the environment. The recognized objects can be used in order to determinate or verify the location of the robot

The Autero multi-robot system

Autero was designed according to the SOA paradigm. The system components communicate in accordance with specified protocols and are as follows:

•           Repository stores  ontology and provides access to it to the  other system  components. It also has  a graphical user interface (GUI) for developing  the ontology and its management.
• Task Manager (TM for short) represents a client and provides a GUI for  the client to define tasks and monitor their realization. Task is defined as a logical formula that describes the required  final situation in the environment by  using types,  objects and relations from the ontology stored in Repository.
•         Planner provides abstract plans for TM, that are used  to construct a concrete plan on the basis of information about available services (provided by Service Registry)  and by arranging these services (by the Arrangement Module) in a business process. TM controls the plan realization by communicating with the services arranged in the plan.  Arrangement is performed by sending requests (in the form of intentions) to services and collecting answers as quotes (commitments). 
•          Service Registry stores information about services currently available in the system. Each service, in order to be available,  must be registered  in Service Registry  via Service Manager.
•          Service Manager  is a  robot interface for providing its services for an external client.It controls the execution of the subtask delegated by Task Manager and reports back a success or a failure.

A Simulation environment

The simulation environment is developed in Unity3D. It serves as a platform for testing and validating proposed specifications of the environment representation (ontology), the architecture of a multi-robot system at the level of IT, and learning algorithms. It allows for implementation of more complex scenarios involving multiple services (devices).