CSC 297 Robot Construction: Summary

Course Summary

For the purposes of this class, a robot is any artifact that moves by itself under some sort of adaptive, (non-human-sourced) control. This covers a lot of ground. BigDog is a robot. So is Predator. So is the automatic transmission in a 1990s automobile. 2013 automobiles are complex, integrated robotic systems, even the ones that can't parallel park themselves or keep to a lane on the freeway. Your CD/DVD/BluRay player is a robot. So is your inkjet printer.

On the other hand, a windup walking toy is not a robot - there is no adaptive control. Most of the mechanisms you see in high-school "robot" contests are not robots either, but mechanical devices remotely controlled by a human. Some of them contain components, such as servo-motors that qualify as simple robotic devices themselves, but the visible "robot" aspects are pure theater. All the interesting behavior is coming from the human controllers.

Segways (and automobiles, and modern vehicles in general) are an interesting hybrid case. Certainly the human pilot is supplying some high-level control. But the vehicles are doing a lot of complex adaptive control at other levels, which is necessary for the vehicles to operate as they do. A person could not easily balance a Segway given only direct control of the motors; that is provided by fast internal "reflexes" Modern automobile engines monitor a dozen or more variables, and continually adjust several parameters to optimize performance. The same is true for automatic transmissions, antilock brakes, stability control, even airbags and seatbelts. In a lot of ways, driving a modern automobile is a lot more like riding a horse than riding a bicycle.

Of course the line is blurrable. We could debate philosophically whether a marble that rolls to the lowest point on a surface is subject to adaptive control or not. As designers however, the pragmatic issue is whether it is helpful to separate control from other aspects of a mechanism.

In this class our goal is to design and construct some cool novel robots. We will work to design both the physical artifact and its control, as the two are intimately connected. Novel means we are not simply assembling a kit into a designated end-product. However, we will certainly incorporate and/or modify existing artifacts into our designs. To some extent, the higher-level components we can use the better. That's how technology progresses after all. We will probably not make our own motors, unless we require some really unusual properties, nor mine our own copper; nor will we likely make our own transistors, or refine silicon. If something we can use is available at reasonable cost, we'll use it. Otherwise we'd not get much further than stone axes and fire, if that far.

Rather than beginning with textbook material, we will start by dreaming up some bots to make that can't be purchased off the shelf. We'll then learn what we need to know to make them. I anticipate 3-4 projects with small teams of ~5 people will working on each. The prof will attempt to keep projects from wandering too far from what is doable.

At the end of the semester I think the most you can expect is some spotty, varied exposure to (the large subject of) robotics, but should you encounter more complete material later on, you may find you have a feel for what about it is important and why. You may also acquire an appreciation of the difference between the elegant theoretical models in textbooks, and the grubby reality of getting these models to be helpful in a real design process (because sometimes, sometimes they are, they're just never the whole answer). That at least, in addition to having some fun, and dealing with many unexpected difficulties, is my hope for this course.

Students will be expected to attend at least one 3-hour group lab a week, and should expect to put in at least 8 hrs/week of additional out-of class time. Building bots ain't for sissies.


Back to CSC 297 main page