1. Introduction and Overview

Students using this research module would focus on the analysis and design of entire, yet simple, autonomous agents. Such a module lends itself toward more bottom-up methodologies than have been employed in the PT project so far, thus allowing PT participants background for understanding the relative merits of top-down and bottom-up approaches. The projects pursued this module would also afford students an oppurtunity to appraise the boundaries between problems that do and do not require representation and computation for their solution.

A unifying source of inspiration for the activities described below is the book Vehicles by Valentino Braitenberg. Braitenberg described, in a series of thought experiments, autonomous agents that exhibited deceptively complex behaviors. The word 'deceptively' is appropriate here becasue the architecture of the creatures is actually quite simple. Each vehicle is comprised of only three basic components: sensors, motors, and sensorimotor connectors. Connectors may be either inhibitory or excitatory. By varying the number and kinds of components constituting that vehicle, the vehicle may be capable of acheiving relatively complex pursuit, avoidance, and navigation behaviors. More information regarding Braitenberg's vehicles may be gleaned in sections 4.1 and 5.1 below.

The first and core research activities that students would pursue would involve experimenting with virtual and robotic autonomous agents to program them to emulate the architecture and behaviors of Braitenberg's vehicles. For example, students would seek to create an agent that successfully navigated through an environment and sought out attractive stimuli while avoiding noxious stimuli. Vehcile designs would be implemented in robotic agents using Khepera miniature robots. Designs would be implemented in virtual agents using either or both the Popbugs and Khepera Simulator Freeware packages.

A second line of investigation would concern creating an interface between the ProtoThinker model of higher cognition and the simple mobile control structures instantiated in simple Braitenberg vehicles. Real persons seem to have separate resources by which they accomplish the separate demands of abstract thought and reflexive obstacle avoidance. Perhaps artifical and virtual persons would best be designed this way too. Thus, the architecture of an obstacle avoiding vehicle might be integrated into IRIS to form the equivalent of our unconsious motor abilities (we can walk without thinking much about it). However, a vehicle/PT hybrid won't just consist ofPT riding on the back of a vehicle, so to speak: there will need to be some amount of communication going both ways. The PT, higher, level will need to know what the vehicle, lower, level is doing ('where am I and how did I get here?'). Likewise, the automatic procedures of the vehicular level must be open to over-riding commands from PT: while walking is largely automatic our higher cognitive centers do have some say-so in where we're going. The design of a PT/vehicle hybrid along the lines just described would instantiate what robotocist Rodney Brooks has dubbed a 'subsumtion architecture' (see Clark 1996 (described in sec 4.3 below) for further discussion of this concept.

A third line of research that students may pursue in this module is the writing of a Graphical User Interface Vehicle Editor. Such an editor would have a palate based interface (as found in a typical-draw-and-paint program) allowing users to select vehicle components from a pallate and building vehicles via a series of mouse clicks and drags. The interface of the finished editor should be simple enough to allow use by very young students (grade and middle school). Thus the commpletion of the Vehicle Editor would comprise and expansion of the PT project into a wider range of age-groups.

The remaining sections of this document are organized as follows. In sec 2 I describe the hardware and software needed to perfom the above activities, especially the first and second lines of research described. In sec 3 [still incomplete] I outline the kinds of projects students would pursue. Section 4 contains brief descriptions of textbooks that I think would be the most desirable sources of background readings. Section 5 contains links to further online resources relevant to the material discussed here.



Previous Contents Next