Hospital delivery go-kart

the system at a glance

The problem

Hospitals are understaffed, and nurses lose time to small errands between rooms. The brief was a compact, modular vehicle that could take some of those runs: delivering food and medication around a controlled indoor space without spilling and without overheating.

Concepts we considered

Each of us sketched a concept and we scored all four on weighted criteria, with safety and function (no spills, no overheating, carries food, reliable) weighted highest:

Concept	Idea	Score
1	Belt drive, PLA frame, joystick control, a secured bed	15.8
2	Chain drive, carbon frame, separate trailer	11.0
3	Planetary gearbox, aluminium chassis, trailer	14.4
4	90° bevel gearbox, steel frame, lockable safe and solar panel	21.8

We merged the two best: concept 4’s enclosed, compact 90° gearbox and concept 1’s lightweight printed bed.

The design

A 12 V XD3420 DC motor drives the rear axle through a bevel pair that turns the drive 90°, then a spur pair. A servo on a simple ball-joint linkage steers the front wheels, and the front has suspension springs. An Arduino Uno runs four control modes:

Joystick: direct analog driving. The stick reads 0–1023 and should rest at 512, but drifts, so we measured the drift and coded a dead zone around centre.
IR remote: decodes each button’s code into drive and steering commands.
Phone app: sends commands over Wi-Fi through Blynk.
Autonomous: an ultrasonic sensor watches ahead; anything within 20 cm stops the kart and it turns away.

Steering is set to three angles: 53° straight ahead, 90° left and 135° right.

My part: layout and gearbox

I owned the embodiment design and definitive layout, and the gearbox calculations. Starting from the motor’s speed at peak efficiency, the target wheel speeds gave two ratios — about 2.1:1 for high speed and about 8.6:1 for low speed — and I laid the gearbox out to shift between them. The calculations went through gear efficiencies for each bevel and spur pair, pitch-line velocities, and the tangential, radial and axial forces on each gear.

The team sized the chassis plates by hand for steel, aluminium and PLA. PLA at 75% infill came out lightest by a wide margin, and FEA confirmed the top plate could carry the electronics.

Testing

Function	Result
Joystick, IR and app control	All worked; the dead zone stopped drifting at rest
Obstacle avoidance	Stopped and turned away from objects inside 20 cm
Steering	Hit 53°, 90° and 135° reliably, with minimal backlash
Two-speed gearbox	Not included in the final build

At the bottom of this page there’s a 3D model of the kart built from the dimensioned layout: you can steer it to the three tested servo angles, run the drive, and look through the gearbox housing at the bevel and spur pairs as designed.

What went wrong

The shifting never made it in. Three problems stacked up:

Bevel geometry. Bevel gears only mesh properly at the exact cone angle, and we had very little experience with them. Small angle errors meant partial meshing, slipping and lost power.
Tolerances. We bumped the gear module from 1.0 to 1.25 to make alignment more forgiving, but printed parts still weren’t precise enough to shift reliably.
Heat. Under sustained load the PLA gears softened, ground down and degraded.

We shipped a single-speed version that drove, steered and avoided obstacles. It was a useful gap to run into: the gearbox was fine in CAD and in the calculations, and the problems only appeared once real parts were meshing. Next time I’d use metal or bought-in bevel gears, check contact ratio and heat build-up properly, and test a bare gear pair long before the full assembly.