The domain of Cognitive robotics tries to recognize manipulation tasks which are demonstrated by humans and other robots. In the simplest form, it's equal to a video parsing system1 but it can be extended with learning capabilities.2
Before a robot can execute tasks, the environment has to perceived with Robotic sensors. The raw data are converted into machine readable information which are enriched with semantic information. Natural language grounding (Symbol grounding problem) is equal to convert the actions of the robot's environment into textual information. Semantic Event Chains 3 and Object-Action complex are used to store the information in a database.
Norbert Krüger and Christopher Geib and Justus Piater and Ronald Petrick and Mark Steedman and Florentin Wörgötter (2011). "Object-Action Complexes: Grounded abstractions of sensory-motor processes". Robotics and Autonomous Systems. 59 (10). Elsevier BV: 740–757. doi:10.1016/j.robot.2011.05.009. hdl:2117/15586. /wiki/Doi_(identifier) ↩
Eren Erdal Aksoy and Alexey Abramov and Johannes Dörr and Kejun Ning and Babette Dellen and Florentin Wörgötter (2011). "Learning the semantics of object action relations by observation". The International Journal of Robotics Research. 30 (10). SAGE Publications: 1229–1249. doi:10.1177/0278364911410459. hdl:2117/14016. /wiki/Doi_(identifier) ↩