Robotics as an Educational Tool

This paper explores a new educational application of Piaget’s theories of cognitive development i.e. the use, as a teaching tool, of physical robots conceived as artificial organisms. By using simple assembly kits, students at all levels are able to project and construct real robots that simulate the behaviors of animals. The process of constructing real robots helps students to understand concepts about complex dynamic systems – in particular how global behavior can emerge from local dynamics. This is done through a construction process. In order to obtain a given behavior students modify both the “mind” and the body of artificial organisms. The construction of populations of artificial organisms helps the students to realize the difference between observing behavior at the individual (microscopic) level and at the population (macroscopic) level. The development of a population of robots with a given behavior is an evolutionary process. The selective reproduction of a population of robots is a powerful tool for teaching the Darwinian theory of evolution: experiments using artificial – as opposed to biological organisms make it possible to rapidly observe the results of selection, reproduction and mutation. The paper reviews a number of educational projects using real robots. It is shown that the use of intelligent systems to enlarge our view of biological reality could become an integral part of curricula in science, technology, psychology and biology. To appear in Journal of Interactive Learning Research ARTIFICIAL ORGANISMS FOR EDUCATION During the last decade, researchers and industries have proposed and developed a number of “robot construction kits” designed to stimulate the learning of concepts and methods related to the education of students in scientific fields such as mathematics, physics, computer science, and mechanics. The kits include small motors, simple sensors, wheels, gearwheels, belts, and relays everything the student needs to construct a robot. Products like LEGO Dacta and LEGO CyberMaster include cables or radio equipment making it possible to connect the robot to a personal computer. This allows a user to control the device. The recent LEGO Mindstorms product has been developed so as to allow the user to build fully autonomous robots with all computing power located within the machine. These kits have been built in accordance with educational principles derived from Jean Piaget's theories of cognitive development (1966) as revised by Seymour Papert (1980; 1986). This approach suggests that the center of all learning processes is the active role of the learner who enlarges his/her knowledge through the manipulation and construction of objects. This philosophy suggests that traditional construction kits are highly suitable for use as learning tools. Giving life to an object through interaction with a personal computer makes it possible, however, to develop applications which go beyond the original ideas of those who first proposed this methodology. In particular, a number of research groups have constructed small, mobile machines that simulate the behaviors of real animals. Such prototypes are essentially mobile robots. Like real animals they have a sensory apparatus (i.e. sensors which are sensible to light or heat), a motor system (e.g. mechanical arms or wheels controlled by motors) and a brain (represented by a computer programmed to control the motor system using information from the sensory apparatus). These machines can be treated as Artificial Organisms and are used both for educational purposes and for fundamental research in fields such as psychology, ethology and robotics. ARTIFICIAL ORGANISMS AND NEW EDUCATIONAL TECHNOLOGIES Artificial Organisms and the Understanding of Complexity The molecules of a gas within a gasometer, the genetic codes of living beings, an organism's brain, the bees in a beehive and human communities are all examples of complex dynamic systems. A system is defined to be complex when it is constituted of many different elements that interact with each other. It is a dynamic system when the micro-laws of interaction amongst the different elements produce macro-effects which vary over time. Scientific interest in complexity has produced more than just technological knowledge; what it has done is create a new way of observing and interpreting reality. This is based on the knowledge that in a complex system each component interacts with other components and thus that any action by one component influences the behavior of other components. As a result global behavior emerges from local dynamics affecting specific sub-systems. External disturbances or modifications in the interaction principles governing the activity of system components leads to changes in these local dynamics. These are regulated by non-linear laws. Small, random fluctuations in the behavior of a single component can produce huge changes in global behavior. At the same time however major disturbances can sometimes be absorbed, leaving the state of the system unchanged. Consequently, to study complex dynamic systems, one has to consider behavior both at the microscopic level (the behavior of single components) and at the macroscopic level (the collective, global behavior produced by the interaction of all components). To transfer this new way of perceiving reality to children, or, more generally, to persons outside the world of scientific research, requires new teaching tools. The importance of the task is evident: we are not just talking about new notions or concepts but about new ways of observing and reasoning that might help people to evaluate the reality in which they live more attentively. Mitchell Resnick, from MIT’s Media Lab, has developed a teaching methodology that allows the learning of concepts essential to the understanding of complex dynamic systems (Resnick, 1988, 1989, 1994; Kafai & Resnick, 1996). This work is an important part of the background to LEGO Mindstorms (see Appendix A). Resnick proposes a work group of pupils who are to construct "artificial organisms". The pupils follow a precise construction plan but have the freedom to introduce variants. A concrete example of the potential of this approach can be found in experiments in which a group of children is asked to construct an artificial organism with the ability to move towards a light source. The first phase in this experiment is to design the body of the machine, i.e. to construct the hardware structure of the robot, decide what kind of sensors to use and how many of them there should be, to define the motor apparatus (choosing wheels, belts or artificial legs). A simple hardware structure for a mobile robot would be a box mounted on two wheels with a light sensor on the front. Each wheel is controlled by an electric motor. In the simplest design, a motor can be on, and thereby provoke a forward movement of the wheel, or off, in which case there will be no movement. In this way, a robot with two independent motors, each connected to one wheel, can produce 4 kinds of actions: go forward (when both motors are on), turn right or left (when one motor is on and the other is off), or stay still (when both motors are off at the same time). The characteristics of the sensors are such that activation is directly proportional to the distance which separates the sensor from the light source. Having constructed the body of the artificial organism, pupils have to give it a "mind". In this phase, pupils program the computer controlling the behavior of the robot. If they want light-approaching behavior, the pupils have to write procedures where motor activation is a function of the intensity of light perceived by the sensors. At this point a discussion arises: how can an artificial organism with only one sensor move towards a source of stimulation? Usually, the children realize that, as with real, living organisms, there are two different solutions to the problem. In solution (a) the robot reads the light intensity perceived by the sensor at two different moments; if light intensity at moment 1 is less than intensity at moment 2, the robot is moving towards the light and the correct action is to carry on forward. A second solution (b) might be to add an extra light sensor on the rear of the artificial organism and sense whether the activation on the forward sensor is higher than activation in the rear, in which case the correct action is again to carry on forward. At this point, the instructor suggests alternative solutions, pointing out that solution (a) requires "memory", i.e. a change in the "mind" of the robot, whereas (b) consists of a structural modification, i.e. a change in the "body" of the robot. In a second experiment children were asked to construct a population of artificial organisms and observe their behaviors at the individual (microscopic) and population (macroscopic) level. The population consisted of two different kinds of artificial organism: one category of robot was programmed to move toward light sources while a second category was programmed to move away from any kind of light. In this way, each individual had a rather simple behavior. If, however, we place a small lamp on the "head" of each organism the behavior changes – in interesting ways. This may lead to one of two alternative patterns of global population behavior. In pattern (a) it is observed that if the two categories of robot are initially segregated into different regions of space, organisms that are attracted to light tend to meet, to bump into each other and to concentrate within a very small area; robots which avoid light tend, on the other hand, to scatter through the environment until each individual is at “a safe distance” from all other robots. In pattern (b) there is no initial spatial segregation; this implies that an individual belonging to one category can interact with individuals from the other category; in this case one observes complicated patterns of flight and pursuit. Between patterns (a) and (b) there exists a large number of intermediate solutions. Practical experiments such as these help learners to assimilate concepts which would otherwise be abstract and obscure. The children assimilate the notions of dynamics and complexity through the construction of systems composed of a number of hardware and software components. They learn to study reality from different points of view (i.e. at different levels of analysis) by observing the behavior of individual robots and the global behavior that emerges from the interaction between these individuals. Artificial Organisms in Undergraduate and Graduate Courses It has been observed that engineers with a bachelor’s degree often have an excellent knowledge of fundamental theoretical concepts in their discipline but have had insufficient experience in designing and constructing industrial prototypes. Over the last four years, Fred Martin of MIT has organized a course on "Designing and Constructing LEGO Robots" (Martin, 1994, 1996), the goal being to stimulate design and implementation capabilities in young engineering students. The students in the experiment were divided into small working groups which were given the task of designing and constructing a mobile robot to solve a task given by the teacher (e.g. to move from one point to another while avoiding obstacles of different forms and dimensions). At the end of the course, the prototypes built by the students participated in a competition; the group which produced the most efficient robot won a prize. The contest has now become an annual event at MIT, attracting strong interest and enthusiasm from the entire MIT scientific community. The Department of Artificial Intelligence at the University of Edinburgh also uses robot competitions as part of its curriculum. In his course Intelligence and Sensing Control, John Hallam lectures on artificial intelligence approaches to robotics. Students are given the Edinburgh brain brick which can be used, like the MIT programmable brick, to control LEGO robots. The competitions include robot sumo wrestling and robot rugby. At the Aarhus University Department of Computer Science we also use robot competitions in graduate level courses on Adaptive Robots and Robot Modelling. In the different competitions, the students are given a Khepera miniature mobile robot; alternatively they build their own LEGO Mindstorms robot. The competitions include the Danish Championship in Robot Soccer. Building a robot allows computer science students to learn about real world applications. Traditional computer science courses seldom teach the students about the uncertainties of real world interaction; usually, in fact, they try to abstract from the real world and build fully deterministic systems. This can cause problems when computer scientists are later supposed to design and/or program real world control systems. Figure 1. Photo from one of the robot soccer tournaments (Copyright 1998. H. H. Lund) While designing robot soccer players, computer science students always make a number of mistakes, usually due to an unrealistic view of their robot’s effective capabilities. These problems are often due to the students’ failing to view the robot from its own point of view, relying on the sort of unrealistic abstractions they have been trained to use for most of their education. Through experimentation with sensors, motors and control, the students gradually modify their view of the interaction between the robot and the real world, continuously modifying their design until it becomes a realistic one. According to Fred Martin the success of this kind of teaching experiment is partly due to the fact that the construction kits available on the market facilitate assembly. Available construction kits allow students to find simple solutions to physical problems. The students achieve a real feel for the discrepancies between the results predicted at the design stage and those actually produced by their machines, learning to reduce this discrepancy during design and construction. In this way, the students become acquainted with the circular relationship between theory and practice that is fundamental to technological innovation. Robotics in High School At the LEGO Lab at Aarhus University, we have used robots with high school students on a number of occasions. Recently, we arranged the FIRST LEGO League as a pilot project for a number of Danish high schools. The pupils were 12-14 years old. Each class was given four LEGO Mindstorms sets. The students worked in groups of 4-5 to plan, design, build, and program their own robot to participate in the competition. The task was focused on building a robot which could quickly navigate an arena with black tracks on the floor, a ramp, small obstacles, etc. There were awards for the best performing robot but also for the most beautiful robot and the bravest robot. Figure 2. Robot as a “real” Pacman. (Copyright 1998. H. H. Lund) The LEGO League pilot project ran for a short period of a couple of months in the autumn of 1998. During that time the classes worked intensively on the project. They started out knowing nothing or very little about robotics, and little about programming (no one knew the Mindstorms graphical programming language before the project started). But, 1 Here we use the English term high schools though these are not high schools in the Danish sense, but schools with pupils from the age of 6 to 15 years old. with enthusiasm and help from tutors, the pupils managed to complete the work and to have well performing robots ready for the final. Boys and girls were both highly involved in the project. For instance, a group of girls decided to build a bride and a groom to run around together. After much work on the aesthetics they wanted to put specific functionality into the robots. In this way they became interested in programming and actually learnt to program the robots, though initially they had just referred to the boys when this kind of work was needed. ARTIFICIAL ORGANISMS AND BASIC SCIENTIFIC RESEARCH. Braitenberg Vehicles Approximately 10 years ago, Valentino Braitenberg, one of the pioneers of cybernetics, published a small, but very interesting booklet entitled Vehicles: Experiments in Synthetic Psychology (Braitenberg, 1984). In the book he suggested – as a provocation – that it might be possible to gain insight into a number of typical psychological research themes by constructing small mobile robots that behaved as if they possessed sophisticated mental states. In the booklet, Braitenberg describes a number of experiments where small, vehicles of gradually increasing complexity are constructed out of simple mechanical and electrical components. Each of these machines in some way imitates an intelligent behavior; each is given a name which corresponds to the behavior it imitates. It is important to emphasize that Braitenberg never actually built real robots, but limited himself to designing and describing the robots on paper. He calls this methodology "Synthetic Psychology". What Braitenberg shows, with his work, is that, despite their simple mechanics, his machines (vehicles), show behaviors that an external observer would classify to be the product of mental states such as fear, embarrassment, hesitation, paranoia etc. Though Braitenberg’s ideas have produced controversial reactions, it is undeniable that they have had a strong impact on basic research, inspiring a broad range of different studies. Using a special kit of programmable bricks, David Hogg, Fred Martin, and Mitchell Resnick from MIT Media Lab have constructed the main parts of Braitenberg’s vehicles (Hogg, Martin, & Resnick, 1991). Lund and Miglino (1995) have produced the same series of Braitenberg vehicles using the basic hardware structure presented in figure 3. Figure 3. General configuration of the hardware of a Braitenberg vehicle. Timid (shadow seeker). The robot has one sensor that senses light intensity. The vehicle moves forward if environmental light exceeds a pre-defined threshold, stopping if it finds itself in a shadow zone (see figure 4). Despite the simplicity of this behavior, an observer will typically attribute sophisticated mental states to the organism (e.g. "looking for shade", "hiding", etc.).