--- PART 2 ---
Now - if that kind of a platform seems too small, there is another option that isn't as dangerous as a full-sized ATV - which is what I am pursuing:
PowerWheels
You can find 6 volt and 12 volt PowerWheels ride-on toy vehicles which don't move too fast, are fairly easy to control, and if they get out of hand they won't likely cause too much damage. Their drive motors are fairly easy to control; the steering system is where it gets more difficult (you'll run into the same problems with an ATV - so starting on a ride-on toy may allow you to solve the problems in a fashion that you can then transfer to the ATV when you get there). Some solve it with geared servos, some solve it with other means. I'm personally going to be using (well - it will be an attempt) a 24 volt linear actuator I got surplus. Incidentally, such a similar actuator could easily rotate the steering of an ATV (it's has 4000 N push-pull rating - that's 800 lbs of force); should I get to that point, I can easily move the component over. There are many forums and information out there on doing "PowerWheels Modding" - many people have made remote-controlled PowerWheels toys, and have documented them on the internet; you might want to look at this possibility. New PowerWheels are fairly inexpensive, and used ones can be found cheaply as well (kids outgrow them all the time). They are very tough, but made out of an easy-to-modify using handtools plastic that makes them very adaptable to homebrew robotics.
The problems with an ATV platform are several fold - you need special tools and skills to "interface" to such a machine. Do you know how to weld? Have you ever used a right-angle grinder? Do you have experience and access to a cutting torch? Do you know anything about vehicle repair and maintenance? Small engines? Metal work? Do you have a place where you could work on such a machine (large garage or outdoor area)?
Not to mention the cost of an ATV (though maybe you already own one, or maybe you know where to get one cheap)...
Much of this can be avoided by going with something smaller, and something inherently safer. I've heard more than a few stories about people using, for instance, a differentially steered 200+ lb mobility chair or electric wheelchair as a robot platform in a similar manner, who hooked it up and something went wrong, and it went careening down a hallway, crashing into a wall (if they are very lucky). Can you imagine what would happen if your code was incorrect, or the mechanical linkages were not sound, and you several hundred pound ATV took off at full throttle? Well - can you? Nothing good, I can easily say that much.
...and this is why - before anything else - before you start modding your R/C car, your PowerWheels, your ATV - or whatever other platform you use - large or small - this is why you give some long and drawn out thought to the general layout of your control system, with particular emphasis on safety and control recovery. DO IT NOW. Don't wait until you have a runaway on your hands and you are filing an insurance claim (or having one filed against you).
You want a system that will allow you control, but knows when things are out of control, and can shut down automatically. You want a system where you can remotely shut it down completely. You want blinking safety strobes and possibly warning sirens. You want an on-board, large flashing red "emergency stop" button that is clearly marked and easily accessed from the outside of the machine. You want a "dead-man" safety-pull switch that can possibly be easily grabbed, and once pulled out, stops the machine quickly (like those used by jet-ski riders and similar). You might even want an automatic fire-extinguishing system installed.
You may think these things don't belong on an R/C car or a PowerWheels - but I can assure you that a few of them do (maybe not all - but more than a few). All of them (and probably more) should be implemented once you move to some form of a larger-scale (and possibly engine-powered) machine, like an ATV. The important thing is to design these systems ahead of time, and implement what you can (and what makes sense) for the scale of platform you do use, and to put these in place FIRST - so that when you transfer your design to a larger system, your safety systems are in place without having to band-aid them in afterward. This is the same kind of thinking and design practice as should be done in computer security setups; you should never set your security systems up last (especially if you don't want to be hacked - sadly, in many situations, the norm is to think about security last - I call this idiot engineering).
Know that despite all of these precautions and all of your design choices, testing and such, you might still end up with a "runaway"; in the case of an R/C vehicle or that of a PowerWheels (depending on what you have on-board), it probably isn't that big of a deal. You can easily stop it with your foot, or run after it or something and bring it to a safe halt. In the case of an ATV (or a larger platform with a lot of weight in SLA batteries!) - running after it might be impossible, or even if you did catch up to it, you might not be able to halt its momentum. If all of your safety systems and such fail - if all of your precautions do nothing; well - first, I hope you are testing your large robot where there isn't much to harm (particularly humans), but even so...
Buy some insurance coverage. This really only applies in the case of a UGV based on something like an ATV or a full-sized automobile; it won't be cheap, it won't be easy - but it will be cheaper and easier than the alternative.
Good luck with your project, and please re-read everything I have written here. It is very important, and shouldn't be taken lightly. This isn't an easy project (even at the level of an R/C vehicle); and once you go past R/C vehicle size, it isn't cheap, either. For my project, using a PowerWheels, I've probably spent $1000.00+ USD in mostly used components - not all at once mind you, but over the course of time. I know that if I get to building an ATV version (which is a goal of mine), I'll probably end up spending 10 times that, and probably more.
