The project R-CASTLE (Robotic-Cognitive Adaptive System for Teaching and LEarning) aims to deliver a new educational tool that allows anyone to easily plan and perform interactive activities with electronic devices.
A NAO robot is being used in ongoing studies to validate this project’s methods, providing autonomous adaptive behavior and natural communication to the users for short and long-term interactions. Deep learning neural networks are implemented for recognizing: objects, users’ faces, and users’ emotions. Verbal answers analysis is made through Natural Language Processing algorithms for Portuguese. Adaptation is shaped with Rule-Based and Fuzzy System methods using a multimodal reading of the users’ body signs as input.
Information from all sessions are stored in the system database and can be accessed at any time through graphs or charts. An automatic overall analysis can be performed by the system for supporting teachers to enhance the design of next activities based on critical points detected in previous interactions.
Results from studies with users have been supporting the hypothesis of learning experience enhancing for both teachers and students that use the system.
Code:
https://github.com/LAR-Educational/Architecture_v2_0/
Technical Scheme
The proposed architecture aims to offer a framework capable to be programmed for controlling the robot’s resources, addressing previously known and new learning contents in interactive educational activities. Summarizing, there are some open-source programmed behaviors (as object and face recognition, web research, speech processing) and an interface where a non-programing designer can use all robots features to create a new activity. A NAO robot was used as the device in our tests due to making part of our lab tools but the architecture is easily adapted to control other devices as well.
The R-CASTLE is divided into modules according to their functionalities: Vision, Dialog, Motor for Interactive Module and Memory, Content, Evaluation and Adaptive functions in Cognitive Module as illustrated in figure below.
Click here to see technical details and unpublished results of R-CASTLE_1.1.
Daniel Tozadore 