The Robbie Next Generation Project

Back

The Robbie NG project is one of many projects the Hogeschool van Utrecht (HvU) has devised for its students. All students have been grouped in project teams and each team has been assigned to a project. This project is part of a co-operation between several faculties of the HvU to develop real robots by students electrical engineering, hardware-interfaces by students computer science, and an operating system by students computer science. The Robbie project team is the team that develops the operating system.

The Robbie NG project team started out as the Robbie í98 project team, which has developed a system written in JAVA to control a small robot. The definition of the Robbie NG project was to replace that system by a complex adaptive system. We were to make robots work as a team, having a collective goal to achieve. The supervising teacher that has been assigned to both Robbie projects is Reinoud Karssenberg, who has extensive experience in Object Oriented modelling and programming. To view the achievements of the Robbie í98 project, you can have a look at http://come.to/robbie98.

A part of the project is to make students aware that a strict scheduling and registration of work is an important aspect of system development. We have created a website on which each team member can register his working hours, and we can all monitor other members progress. We have no team leader, the responsibilities for the direction of the project are carried by each member.

The first phase of the project was to investigate the possibilities for modelling a complex adaptive system. Information about this is hard to find, since it is a very recent concept. We use Rational Rose for modelling our project and agreed to create a model in which knowledge-bases are represented as classes. Another part of the investigation was how to implement a knowledge-base in JAVA. A very useful package was found, the Jess-engine. Further information can be found at the links page.

An important issue with developing a complex adaptive system is what collective goal it should have. Several interesting ideas have crossed our minds, but finally we decided to make the robots play hunter and prey. Other ideas were for example having them search for food, follow a leader, or move as a group.

During the project we found out that knowledge bases may not contain functionality that should be implemented in the application itself. If it does, it will be harder to give the robot a new goal by simply replacing the knowledge bases. It is too late for this project group to correct this error, but we learnt that this situation needs to be avoided in future projects.

We had to be able to simulate the robot system and for that purpose a simulation package was developed, parallel to the project. An argument to develop a simulation is that when this project is finished, we canít tell whether there will be enough real robots available to test the system.

The Robbie NG project is the last one that can be finished by this team. In the future, other teams of students wil continue this project. There will be a time that the classrooms of the Hogeschool van Utrecht are infested by intelligent man-made robots. It is a pity we may all have graduated at that time.

Back