CRESTA: A Cognitivist Robot Execution framework for Semantic-driven Task Awareness

Overview

CRESTA is a skill-based robot control framework addressing Task Awareness built on top of ROS. It brings together:

• Explicit knowledge representation: i) uses a custom world model so that all robot capabilities, object types, and world relations are openly encoded; ii) skills are parameterized, with human-readable pre-, hold- and post-conditions.
• Layered, hybrid deliberative–reactive control: i) at the top level it automatically generates a PDDL planning domain and problem from the current world model, calls a planner, then manages the resulting plan for execution; ii) at the lower level, reactive primitives (e.g. motor commands, sensor feedbacks) tick at high frequency, checking pre-, hold-, post-conditions and allowing fast preemption or failure recovery.
• Modularity and flexibility: the architecture aims to be easily configurable, and its base modules aims to be expandable with custom plugins for ontology, reasoning, perception, and low-level control.

The repository contains a package for each component of the system (plus some utilities):

  root/ 
  ├── README.md 
  ├── CRESTA Task Manager (cresta_manager)
  ├── CRESTA World Model (cresta_world_model)
  │ ├── Ontology
  │ │  └── Plugins 
  │ └── Reasoner 
  │ │  └── Plugins
  ├── CRESTA Perception (cresta_perception)
  │ └── Plugins
  ├── CRESTA Action Awareness (cresta_action_awareness)
  ├── CRESTA msgs (cresta_msgs)
  └── .gitignore 

Requirements

To use CRESTA you must have ROS installed on your machine (choose the correct branch according to your ROS distribution). You also need pip to install python dependencies:

• Unified Planning ( AIPlan4EU Unified Planning )
• JSON

Task Configuration

TODO

How to launch CRESTA

The user should launch his/her plugins developed, which are linked to the framework and configured in the task configuration files (see Task Configuration).

Each component package has its own launch files to be executed (you can use roslaunch or mon launch). If you installed the requirements in a python virtual environment, remember to insert it into the PYTHONPATH:

export PYTHONPATH=/home/<user>/<path_to_virtual_environment>/lib/python3.8/site-packages:$PYTHONPATH

The following is a list of the main ones:

- mon launch cresta_world_model cresta_world_model.launch 
- mon launch cresta_world_model cresta_reasoner.launch  
- mon launch cresta_world_model cresta_ontology.launch  
- mon launch cresta_perception cresta_perception.launch  
- mon launch cresta_manager cresta_manager.launch   
- mon launch cresta_action_awareness action_awareness.launch

For communicating the task name to the Task Manager:

rosservice call /tttask_manager/tttask/set_tttask "tt_task:
	task_id: '<task_id>'
	task_name: '<task_name>' 
	success: 0"

Plugin Configuration

TODO