If you're developing a balancing scooter, you will fall quite a few times. I didn't fall or get hurt at all testing at low speeds, but testing at high speeds earned me a few bruises. Testing such a scooter is not for timid, frail, or un-coordinated folks. Segways look and feel quite safe, but if you build your own balancing scooter, until the software is tuned perfectly it is very likely to lurch violently and throw you off. Controlling and/or leaping off a wildy lurching scooter requires a lot of physical strength and agility and is inherently hazardous. I would rate it more hazardous than skateboarding, but less than racing motocross bikes. Skateboarders fall a lot, but at least they don't have an 80 lb thing under them with spinning tires.
For initial testing at low speed, the most important things are a very handy kill-switch, sturdy shoes and shin guards. Mine, at 80 lbs, isn't heavy enough to hurt much running over your toes, but if you're standing behind it and it backs into you the bare metal plate can really bruise your shins. You also want to make sure to limit the maximum motor speed, so you don't find yourself suddenly testing at high speed.
Unless you have the feedback loop parameters just right, it will start oscillating back and forth out of control when you turn it on. You'll want to shut it down quickly, and the right kind of kill switch is essential. I made a mistake here that you should avoid. I first used a circular power jack of the same kind that most wall transformers use as a combination kill-switch and dead-man's switch. (Terminology: a kill switch is something you press. A dead-man's switch is something that is triggered when you fall off.) I connected the wires together and tied the wire to my belt. Completing the circuit through the power jack turned on the scooter. This is unsafe. When I fell off, it pulled at an angle and ended up ripping the wire instead of pulling out the jack. The exposed ends ofwire touched, and it kept going.
Then I changed to a proper combination dead-man's- and kill-switch made for snowmobiles and jet skis. It can handle being yanked out in any direction, and it has a big red button to press to shut it down manually.
Testing at high speed is scarier. The dynamics of limiting speed are quite complicated and hard to get right, and it's really frightening when it starts lurching around near maximum speed. Ideally you'd test on some kind of springy surface, like a rubberized jogging track. I didn't have one handy, so I tested in a parking lot. It can fail either by tipping forwards or backwards . When it tipped forwards at high speed, I ended up straddling the vertical bar, still holding the handlbars, with my feet on the ground. Once it tips that far the software locks up the rear wheels, so I was able to bring it and myself to a stop without losing balance. In this case, it's better to hang on to the handlebars and use the scooter to keep your balance than to try to get your legs over the bar and risk getting tripped up or run over. You want sturdy shoes for this: real running shoes, tightly laced, or they'll come off.
Tipping backwards is actually worse. One time it started oscillating at medium speed and zoomed out from under me. I landed on my butt and elbows. You really want to be wearing elbow pads when this happens.
I should point out here that I while I was testing experimental software, the motor driver was a solid, reliable product. Testing a motor driver of your own design is much more dangerous. Motor drivers have a tendency to fail in full-on mode when they get overloaded. Imagine getting up near top speed and all of a sudden one of the wheels goes full forward or full reverse. In most designs, not even the kill switch will shut it down. Please, use a dependable motor controller, and test it under some a heavy simulated load before using it.
I never had much trouble with steering and steering control on mine, and never felt in danger of tipping sideways. I don't know what I'd do if a software problem caused a major steering change. In general, you never want more than 7% differential speed between the wheels.
If it turns too fast and you tip off sideways, you need to let the scooter go rather than get yanked around by it. You're more likely to let it go if there aren't cars parked all around. So you really want to find the biggest possible space you can so you'll feel free to just let the scooter go if it goes crazy. If the kill switch is working properly, it should stop pretty quickly.
I haven't gone fast on a slippery surface and skidded out, but I imagine this would be pretty scary. This is the sort of thing that the Segway folks have probably put a lot of R&D into, but isn't practical for an amateur to figure out.
I mention various wacky vehicle ideas on my page. Don't start with them. For one, count on crashing it a dozen times before it finally works. So start with basic metal pieces. Also, it's really important that you can leap off quickly when things go wrong. Something enclosed, like a Roman chariot, would be a lot more dangerous.
The testing cycle is roughly like:
You're likely to have to repeat the cycle 50 times or more to get it working well. In order to make it most convenient, you should really have a laptop you can take with you outside, rather than dragging it up and down the basement stairs dozens of times.
To summarize, you'd really be a fool to test with less than the following. For low speed:
and for high speed: