Pedagogical Agents for Teacher Intervention in Educational Robotics Classes

[ Project | Current state | Publications | People ]



Educational robotics is successfully used for teaching in several school contexts. These small-scale computerized teaching tools have many advantages over the PC based tools which have traditionally been used to teach, for example, programming or engineering. Lego Mindstorms is one well-known example of educational robotics. This robot construction set is flexible and simple enough platform for building and programming even for novices.

Educational robotics sets let room for student's own creativity by emphasizing active learner as the center of the learning process. However, in a typical classroom setting, especially at the elementary level, a teacher might have 30-40 children to teach. When using educational robotics in large classroom settings, students are usually divided into groups of 3-4 students. A typical educational robotics project follows an iterative cycle of building, programming, testing, and evaluation. It is characteristics that groups proceed differently, being in different phases of the cycle at the same time. This causes difficulty for the teacher to notice the needs for intervention. Our approach is to use educational agents to help the teacher to focus his/her attention in potential problems. The problem can be generalized as follows: How could the robotics environment inform the teacher what students are doing and how they are progressing? In this project, we are developing an educational robotics environment which aims to support the teacher to focus his/her attention in potential problems in the classroom.


The implemented system will contain four separate parts that implement independent, agent-like behaviour (Figure 1). The first agent module is an intelligent agent which inhabit in the IPPE programming environment (the IPPE Agent). The main purpose of this agent is to observe the user's activity with the programming environment and build decisions based on the input data coming from the graphical user interface of the programming environment. The second agent module implements the similar observing behaviour than the IPPE Agent, but it inhabits in Lego robots' RCX unit (the LM Agent). Due to the different host platform, we handle these agents as s separate parts of the model.

General architecture
Figure 1. General architecture of the agent environment.

Third agent is a screen-based model of classroom setting with the tools provided by Empirical Modelling environment. The model works as a visualization agent in our system. Finally, we will implement an interaction agent, which has the ability to move from one computer to another, for example from a students' computer to the teacher's computer to report a learner problem observed by the other agents.


The implementation of the architecture has been done mostly with Java-based tools. Java was a natural choice as an implementation platform because it supports threaded applications and communication through sockets, and it this way, provides support for agent's autonomy. Also event-processing mechanism of Java provides good standpoint for the implementation of the agents. The IPPE programming environment and Lego RCX unit's programming platform LeJOS are Java-based environments. Besides Java, the Empirical Modelling tools have been used for the implementation of visualization agent and teacher's GUI. However, due to the fact that EM tools have limited capabilities to communicate with the other applications, we have implement also a communication module as a part of teacher agent. This part will be implemented as a Java application to make the discussion with other agents as easy as possible. Figure 2 presents the general implementation idea and how different parts of the application discuss with each others.
Implementation in general level
Figure 2. Implementation.

Current state

Currently, we are finishing implementation of the first working protoype of the environment. The implemented prototype system has revealed the strenghts and the weakness of the proposed architecture, and it has guided us towards the next version of the prototype which can be tested in real educational robotics classes during the first quarter of 2007.