CoreSense design

CoreSense is a hybrid cognitive architecture that gives robots the capability of understanding, the world and themselves, and of awareness, understanding in synchrony with perception. The theory behind it is described in the public deliverables D1.3 Theory of Understanding and D1.4 Theory of Awareness. The terms used in the software follow the CoreSense Ontology (CSO).

Architecture subsystems

The architecture has three subsystems:

  • Runtime system: a modular software structure, based on ROS 2, used to deploy, monitor and control the modules of the robot. It uses the instrumentation of the ROS 2 platform (see CoreSense Instrumentation).

  • Understanding core: computes meanings on demand, when another subsystem asks for it, for example to generate an explanation for a user.

  • Awareness system: a cognitive structure that continuously uses the understanding core on the flow of percepts. When the object of perception is the robot itself, it generates self-awareness.

The architecture is specified in D2.2 Specification of the CoreSense Architecture, and its software assets are described in D2.3 CoreSense Architecture.

Cognitive modules

A cognitive module is a set of ROS 2 nodes that performs a cognitive function. In the ROS 2 implementation of the architecture, cs4home_architecture, each module has five components, managed through ROS 2 lifecycle transitions:

  • Afferent: receives information from the system or the environment.

  • Core: performs the cognitive function.

  • Efferent: sends the results to other modules or to the robot.

  • Meta: provides information about the module.

  • Coupling: coordinates the module with other modules.

information --> Afferent --> Core (cognitive function) --> Efferent --> other modules
                               ^         ^
                              Meta    Coupling

A master/flow mechanism coordinates several cognitive modules. The tutorial Cognitive modules with cs4home_architecture shows how to build and run example modules.

Understanding system

The understanding system, coresense_understanding, generates strategies to obtain models with given properties. It combines existing models with model-modification skills, the engines. Each engine is a ROS 2 node, annotated with its inputs and outputs, whose skill is wrapped in a behaviour tree. The logic of the understanding core is in understanding-logic, and it uses the Vampire theorem prover through coresense_vampire and the knowledge base triplestar_kb.

Engineering toolchain

CoreSense systems can be designed with model-based tools. See CoreSense Toolchain.