In a world where the population is aging, products that improve living comfort will have more importance in people's lives. These products must interpret the intentions of those who live in the house to provide them with assistance in their daily tasks. Motivated by these issues, we present an architecture for real-time Intention Recognition. We demonstrate it with a kitchen use-case, where the agent prepares a meal. Our goal is to recognize what type of meal the agent intends to prepare. The architecture consists of two layers, namely the “Classification Layer” and the “Problog Layer”. The Classification Layer recognizes the environment through sensors and classifiers, and passes the information to the Problog Layer, which uses Problog to infer the intention. The Problog Layer consists of two Knowledge Bases: the “Static KB” and the “Dynamic KB”. The former axiomatically describes the intentions we want to recognize, while the latter is generated at runtime using information from the Classification Layer.
Predicting humans: A sensor-based architecture for real time Intent Recognition using Problog
D(')Asaro, F. A.
;
2021-01-01
Abstract
In a world where the population is aging, products that improve living comfort will have more importance in people's lives. These products must interpret the intentions of those who live in the house to provide them with assistance in their daily tasks. Motivated by these issues, we present an architecture for real-time Intention Recognition. We demonstrate it with a kitchen use-case, where the agent prepares a meal. Our goal is to recognize what type of meal the agent intends to prepare. The architecture consists of two layers, namely the “Classification Layer” and the “Problog Layer”. The Classification Layer recognizes the environment through sensors and classifiers, and passes the information to the Problog Layer, which uses Problog to infer the intention. The Problog Layer consists of two Knowledge Bases: the “Static KB” and the “Dynamic KB”. The former axiomatically describes the intentions we want to recognize, while the latter is generated at runtime using information from the Classification Layer.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.