Why six wheels
A fixed axle transfers the ground into the deck. An independent corner absorbs it. With six corners you can lose contact at one and lose nothing that matters.
Most ground robots are four-wheeled because four is the cheapest number that stands up. It is a reasonable answer if the surface is reasonable. Ours usually is not.
The case that decided it for us is mundane: a broken kerb with a step, taken at an angle, with a loaded deck. On a four-wheel platform with a rigid axle, that step goes into the chassis, the chassis rotates, and whatever is on the deck rotates with it. If the payload is a rack of temperature-held vials, that is not a ride-quality problem. It is a spillage.
Six independently sprung corners change the arithmetic in three ways.
Contact redundancy. With six wheels you can lose ground contact at one corner and still have five patches driving. The vehicle does not need every wheel to be useful, which means the suspension can be allowed to travel far enough to actually follow the terrain instead of being held short to keep everything planted.
Load per patch. Six patches carrying the same mass means each one carries less. On soft ground, loose fill or wet earth that is precisely the difference between crossing and sinking. It is also why the platform holds its grade figure on surfaces that stop a heavier-per-patch four-wheel machine.
No single mechanical path. Drive is in the hub. There is no propshaft and no differential, so there is no one component whose failure ends the mission. Losing a corner derates the speed envelope. It does not strand the vehicle in a traffic lane.
The cost is real: six drive units, six sets of bearings and seals, six things to inspect, and a harder control problem. We took it because the alternative was a vehicle that only works where the ground is already good — and a vehicle that only works on good ground does not need to be autonomous. It needs to be a trolley.