I'd like to put some numbers to emphasize some of what Andrew and others are saying about torsional rigidity and frame decisions helping drive the live axle vs. dead axle design choice. They have already stated why torsional rigidity is important to consider in chassis design, especially skid steer design.
I'm going to use the term torsion constant, J, to compare the torsional strength of different cross-sectional shapes commonly used in FRC frames. This term defines the angular deflection of a beam about its axis when a torsional stress is applied. In general, open cross-sections like angle and c-channel pieces have less torsional rigidity with respect to closed cross sections like square or rectangular tube. This is easy to discover on your own: you can easily twist a piece of KoP c-channel with your own two hands. If you try to do the same to 1" x 1" x 1/8th" square aluminum tube, you will likely hurt your hands before you observe noticeable angular deflection.
Based on these two sources (
http://www.colemetals.com/channels.htm,
https://engineering.purdue.edu/~ahva...sion_Guide.pdf), I am able to compare the torsion constants of the different cross-sections given similar sizes and the same material.
High Strength Steel (HSS) 2.5" x 1.5" x 0.25" wall rectangular tube has J = 1.14 in.^4
HSS 3" x 1.498" x 0.258" thick c-channel has J = 0.04 in.^4
HSS 2.5" x 2" x 0.25" thick angle has J = 0.023 in.^4
It is impossible (or at least impossibly difficult) to bend sheet metal pieces into a closed cross-sectional shape. This is why most sheet metal chassis designs use c-channel and angle shapes quite often. In comparison, a chassis designed with rectangular/square tube as a main structural element has closed cross-sections. This grants inherent torsional rigidity. Based on the numbers above, a rectangular tube cross-section has about 29 times as much torsional rigidity as a piece of c-channel of comparable size. Wouldn't it be cool if there was a simple way to make sheet metal chassis more rigid in torsion without much added material?
That's where dead axles can come in handy! A dead axle can be implemented in such a way that it turns the cross-sectional shape of the 2 c-channel side rails into an I-beam-with-extra-flanges cross-section over a short distance. This solution also happens to achieve this result exactly where the wheels are supporting the weight of the robot and making it move. Dead axles can be designed to add structural strength to the chassis in places where it is needed most. It can be more difficult to do this with live axles because torque transmission, dynamic loading of the shaft, and shaft retention must all be accounted for as well. If anyone can satisfy all of these design requirements and still provide a torsional member in their live axle, then kudos. But I know for a fact that my team would be taking a huge risk in doing so.
This is why it can often be advantageous to use dead axles in a sheet metal chassis. My team does not have access to the necessary resources to properly cut out and bend dozens of sheet metal parts for a sheet metal chassis. We do have a few small CNC mills, a large (relative to our CNC machines) bridgeport 3-axis mill, and a nice lathe in our machine shop in addition to a mentor with more than 20 years experience welding aluminum FRC robots (how many teams have that???). Based on our resources, it makes sense to use welded rectangular/square tube aluminum for our frame.
Since we do not need added torsional rigidity for our frame members, live axles through the frame members become an attractive option. Luckily, we have the proper machinery to pull off a live axle design (nice lathe + mills). Live axles allow us to more easily cantilever the wheels because the sprockets driven from the gearbox can be run on the inner side of the frame member and transmit the power to the wheels on the outside through the live axles. Cantilevering wheels can be preferable because it allows for easy access to the drive system for wheel, axle, and chain/sprocket issues. Driving cantilevered wheels via chain/sprocket with a dead axle setup is more difficult because the outside of the rail is also a convenient placement for our vertical bumper supports, which would get in the way of the chain.
Live axles also allow us to direct drive the dropped-center wheels of our skid steer drive straight from the gearbox. Even if every link of chain on the robot broke, our drive train would still be in business. In my opinion this single feature heavily tips the scale in favor of live axles because reliability is my number one priority in a drive system. It simply has to work. However, this is just an opinion and others may have a just as valid opinion that dead axles are the way to go because in a sheet metal chassis they provide a method for making EXTREMELY light weight and effective designs.
I am not saying that live axles necessarily correspond to square/rectangular tube chassis construction or that dead axles necessarily correspond to sheet metal designs. I am providing evidence that supports common trends and design decisions in FRC chassis with respect to frame and axle setup. It is up to individual teams to decide what style of chassis/axle setup is best for them based on their resources and expertise. Ironically enough, for far more teams than realize it the clear answer is that the KOP drive train or KBoS is optimal selection.