Legs failed because every joint had to hold up the weight — 8.6 N·m static, 13–17 N·m peak. Wheels put the weight on an axle and a bearing. The motor only has to push it: 1.5–12% of the weight on flat ground. That one change drops the cheapest working carrier from “impossible under $300” to $180.
Used hoverboard donor — hacked with free FOC firmware — on a steel-angle / plywood deck with two swivel casters. Phone-controlled.
✓ Under $300 — salvaged / import-direct only
No credible all-new, US-retail, job-site build — mud, curbs, 20° grass slope, 60 minutes — fits under $300.
Shipping is not in any nominal total. It’s the main reason the sub-$300 stacks are fragile. What drives cost isn’t motors — it’s battery + charger ($40–130), drivers + control ($25–65), wheels/tracks ($40–110), and frame/fasteners ($40–70).
Four complete stacks. No total omits controller, driver, battery, charger, or fasteners. Items marked est. in the lists had no single citable US listing on 11 Nov 2026 — they’re honest estimates, not quotes.
Same formulas as the research, live. This is why wheel size and slope matter more than motor brand — and why a motor whose stall torque equals the number below is still not enough. Work at ≤40–50% of stall.
Crr used here — conservative values for small, low-pressure pneumatic wheels: hard 0.015, gravel 0.040, grass 0.070, mud 0.120. Bicycle calculators quote 0.002–0.02 for big high-pressure tyres — those do not apply to 6–10-in robot wheels that sink and scrub.
Size for ≈5 N·m per wheel continuous on 8-in wheels (2WD, 10° grass) and ≈8–10 N·m peak (20° grass / mud / starting on a slope). So the target is stall ≥10–13 N·m per wheel for 2WD, or ≥6–7 N·m for 4WD.
That’s exactly what a goBILDA 5203 99.5:1 (stall 13.1 N·m) or a geared MY1016Z (≈15–20 N·m output) provides — and what a Pololu 37D 100:1 (stall 3.3 N·m) does not, even in 4WD on a slope.
Acceleration adds more on top: 0.5 m/s² on an 85-lb vehicle is +19.3 N, about +1 N·m per 2WD 8-in wheel. Starting on a slope is the worst case.
Reference case throughout: 35-lb vehicle + 50-lb payload = 85 lb gross (38.6 kg, 378 N). Four weight cases were carried through the research — light wheeled 75 lb, tracked/heavy 100 lb, wheelchair donor 140 lb.
| Surface (85 lb, flat) | Crr | Force | 10° slope total | 20° slope total |
|---|---|---|---|---|
| Hard concrete | 0.015 | 1.3 lb / 5.7 N | 16.0 lb / 71 N | 30.3 lb / 135 N |
| Gravel | 0.040 | 3.4 lb / 15.1 N | 18.2 lb / 81 N | 32.5 lb / 144 N |
| Grass | 0.070 | 6.0 lb / 26.5 N | 20.7 lb / 92 N | 35.0 lb / 156 N |
| Mud | 0.120 | 10.2 lb / 45.4 N | 25.0 lb / 111 N | 39.3 lb / 175 N |
Slope dominates everything. Grade force alone at 10° is 14.8 lb on the 85-lb vehicle; at 20° it’s 29.1 lb — before rolling resistance. A person can push a car on flat ground for the same reason these flat numbers are tiny: rolling force is only 1.5–12% of the weight.
Grass: 27 W mech flat · 92 W mech / 132 W elec at 10° · 156 W / 223 W elec at 20°.
Mud: 45 W flat · 111 W / 159 W elec at 10° · 175 W / 250 W elec at 20°.
Hard: just 6 W flat — but 135 W / 193 W elec at 20°.
Halve at 0.5 m/s, ×1.5 at 1.5 m/s. A 250 W motor per side (500 W total) has ample headroom; 100 W per side is marginal on a 20° grass slope.
Average draw isn’t the slope peak — a job-site duty cycle is mostly flat/grass with slope bursts:
Light flat (25 W avg): 25 Wh usable for 60 min
Mixed grass (60 W avg): 60 Wh usable for 60 min
Rough / slope-heavy (120 W avg): 120 Wh usable for 60 min
Real packs: 2× 12 V 7 Ah SLA ≈ 84 Wh usable (60 min mixed, 30 min rough) · hoverboard 36 V 4.4 Ah ≈ 134 Wh usable · 2× 12 V 20 Ah LiFePO4 = 512 Wh nominal (≈230 Wh usable each).
Why tracks don’t win on torque: use the sprocket pitch radius, not track length. A 3-in sprocket on the 85-lb vehicle needs only 1.76 N·m per track at 10° grass — but add 1.5–2× for skid-steer scrub in a turn plus internal friction, and you budget ≈5–8 N·m per track in practice. Tracks win on ground pressure and obstacle climbing, not torque.
| Mode | Terrain | Turning | Verdict |
|---|---|---|---|
| 2WD differential 2 driven + casters | Flat ★★★ · gravel ★★ · grass ★★ · mud ★ | Zero-radius with casters | Cheapest credible Lowest complexity & cost. Curb limit ≈ ½ wheel diameter. Keep weight over the driven axle. |
| 4WD skid-steer | Flat ★★★ · gravel ★★★ · grass ★★★ · mud ★★ | Skid — drags tyres, 1.5–2× current, tears grass | Cheapest job-site Torque split 4 ways halves per-wheel need. This is what the wheelchair donor gives you. |
| Tracked | Flat ★★ · gravel ★★★ · grass ★★★ · mud ★★★ | Skid, worst scrub; throws loose tracks | Best terrain, worst price Climbs ~⅓ track height. No cheap retail kit exists — DIY belts only under $100. |
| Mecanum / omni | Smooth floor only — rollers jam & sink outdoors | Strafing, elegant | Disqualified A smooth-floor drivetrain. Not a job-site answer at any price here. |
| Hybrid wheel-leg / rocker-bogie | All ★★–★★★ | Good | Out of scope Linkages + extra actuators. Nowhere near cheapest. |
Trap to avoid: sellers quote motor-shaft torque before the gearbox. Cross-check any geared-motor claim with T = 9550 × P(kW) ÷ rpm. A 250 W motor at 120 rpm is 19.9 N·m ideal, ≈15 N·m at 75% efficiency — no matter what the listing headline says.
| Motor | Torque / speed | Price (11 Nov 2026) | Verdict |
|---|---|---|---|
| Hoverboard hub, 6.5-in, 36 V, 250–350 W ea. | 10–15 N·m peak at 15 A · 5–10 N·m continuous · wheel is the motor | ~$10 used / $15 eBay / $35 new ea. | Best torque per $ Needs FOC driver (hacked stock board). Solid 6.5-in tyre is the rough-ground limit. |
| MY1016Z geared, 24 V 250 W, 120 rpm | ≈15–20 N·m rated output (calc.) · ~13 A rated | $38 ea. import-direct · $86–119 US retail | Best new buy 2 needed. The $293 Stack B total lives or dies on the $38 import price. |
| goBILDA 5203 Yellow Jacket 99.5:1, 12 V | Stall 13.1 N·m · 60 rpm · 9.2 A stall · encoder incl. | $54.99 ea. | Honest spec 60 rpm = 0.64 m/s on 8-in wheels — slow. 71.2:1 (84 rpm, 9.2 N·m stall) is the faster pick, same price. |
| Wheelchair motor + gearbox (salvaged) | 250–320 W class, transaxle gearing | In donor · $367–1,186 new — never buy new | Donor only Comes with wheels + controller in Stack D. |
| MY1016 high-speed (no gear), 24 V 250 W | 0.86 N·m · 2,750 rpm | $34.99 / $20.99 direct | Needs ~25:1 reduction The external reduction costs more than the saving vs the Z version. |
| Pololu 37D 100:1 / 131:1, 12 V | Stall 3.3 / 4.4 N·m | $60.95 / $38.95 | Too weak, too dear 4× in 4WD = 17.6 N·m stall total vs ~16 N·m needed at 20° grass — before any margin. |
| Power Wheels RS550 + plastic gearbox | ≈5–8 N·m/wheel peak (inferred from 130-lb vehicle rating) | $0 salvaged · $3–15 replacement | Donor only Plastic gears strip under shock load on rough ground. |
| E-bike geared hub, 36 V 250 W kit | 35–43 N·m max | ~$100–220+ kit | Wrong form factor Huge torque, but bicycle-size wheel, fork mount, kit price. |
| Cordless drill / wiper motor (salvage) | Drill: 20–40 N·m at chuck in low gear · Wiper: 6.2–15 N·m stall, 30–60 rpm | $0–20 junkyard | Fun, not repeatable Custom mounts, no encoder, intermittent duty (wiper). |
Reduction options: built-in planetary ($0 extra, 85–95%/stage) always wins. #25 chain ≈$10–15/side at 95–98%. Timing belt similar, quieter, slips under overload. Worm gear is 50–70% efficient — it holds on a slope for free but roughly doubles battery need; avoid for drive. Sources: Lunyee MY1016Z · goBILDA 5203 · FFBeast hoverboard data · Pololu 37D specs
$49 eBay (Jetson, used, no charger, as-is) · $50 Facebook · $80 with charger. Includes 2 hub motors + tyres, 36 V battery, FOC-hackable mainboard.
Trap: dead/weak battery is the rule, not the exception ($40–150 to replace). Board variant matters for the hack — check the Gen2.x wiki before buying. No suspension, no cargo bed.
Free listings exist (Invacare Pronto M51, no batteries; Permobil, needs battery — Missouri AT, Nov 2026) · $100–500 private · $950–3,500 dealer used. Includes motors, 12–14-in wheels, frame, joystick + controller.
Trap: batteries almost always dead — $200–400/pair as wheelchair SLA, avoided in Stack D with generic LiFePO4. Chair weighs 60–100 lb. Test the joystick before paying.
$20 as-is (“did not charge”) · $75 with new battery · $85–100 working. Includes chassis, 2 motors + gearboxes, 4 wheels.
Traps: plastic gears, dead 12 V SLA ($20–40), hard plastic wheels with poor grass traction, toy steering (remove it, go differential). 50 lb + vehicle on grass is near its design limit.
Also checked and mostly rejected: old mower (wheels + steel deck only — petrol transaxle isn’t reusable), wagon/dolly (you’re buying a trailer, not a drivetrain), dead e-bike/scooter (good for the motor alone; battery usually dead, hub needs a fork). Donor sources: eBay hoverboard listing · MO AT swap listings · used wheelchair guide
2× 12 V 7 Ah SLA (24 V): $38.99/pair · 168 Wh nominal, ~84 usable · ~10 lb
12 V 20 Ah LiFePO4: $52.99 ea. · 256 Wh, ~230 usable · 5.4 lb ea.
Hoverboard 36 V 4.4 Ah: $0 if donor pack is healthy · $40–150 replacement
Chargers: LiFePO4 14.6 V 5 A $13.99 / 10 A $19.99–24.99 · 24 V SLA ≈$20 est.
Never charge Li-ion/LiFePO4 with an SLA charger. Never charge a used hoverboard pack unattended — check for swelling first. LiPo is not recommended here: puncture + job site + fire risk. Drill packs work if you already own the platform (~90–100 Wh each).
BTS7960 / IBT-2: $6.99 ea. (one per motor) — the default. 43 A peak marketing, ~20 A continuous with airflow.
ESP32 DevKit: $9 — Wi-Fi + Bluetooth, phone joystick is $0 extra.
FlySky FS-i6 + receiver: $49.99–52.92 — add if you want a real RC link that works with gloves.
Hoverboard stock board: $0 in the donor, free FOC firmware, needs a ~$10 ST-Link to flash.
Person-following: don’t budget autonomy into the $300 build. Ultrasonic/ToF follow ($10–30) fails in sunlight and follows any obstacle. The credible outdoor method is UWB — 2× DWM1001 at $22–28 ea. (~$50 total) — added later.
SteelWorks 1½-in × 36-in steel angle, $13.99 at Menards ($12.45 after 11% rebate — store credit, not instant). Two pieces make the perimeter; bolt it, no welder needed. Plywood deck ½-in 4×8 is $30.90 — a carrier needs ~¼ sheet ($15 allocated).
Plywood is fine as a deck on a metal frame. It is not the load path for axles or motor mounts.
Harbor Freight 10-in pneumatic, $9.99 ea., 300-lb working load, knobby tread — the default wheel. 8-in pneumatic swivel caster $16.99 ea., 280-lb rated.
Load is 21–35 lb per wheel on 4 wheels. Don’t use printed drive wheels (hub torque shears them) or hard toy wheels on grass.
Tote / mixing tub / milk crate bolted to the deck: ~$12 est. Keep the 50 lb low and centred over the driven axle — a high, rear-biased load unloads the drive wheels and kills slope traction.
Printed parts carry no load here. Resin is brittle; FDM creeps and delaminates. Prints are fine for covers, sensor mounts, chain guards, battery boxes, track guides — never motor mounts, axles, or bed supports.
goBILDA Bravo tracked kit — $1,049.99: the only retail kit found with an explicit 50-lb rating (2 mph, battery + charger + RC included). DIY tracks at $414 are 2.5× cheaper because you are the factory.
SuperDroid IG52-DB4 4WD — $1,674+: 50–75-lb payload, 70-lb vehicle. HD2 tracked: $5,692, 100-lb payload.
No sub-$300 retail kit with a verified 50-lb payload on rough ground was found.
• Pololu 37D motors — too weak, too expensive
• Mecanum wheels — smooth floor only
• New wheelchair motors — $367–$1,186 each
• A retail tracked kit — $1,050+ on this budget
• Mini toy track kits — rated 1.5–2 kg, not 50 lb
• Printed load-path parts — metal or nothing
Hoverboard donor at $180, phone-controlled. Carries 50 lb at walking pace on flat, gravel, short grass. Won’t do mud, curbs, or a 20° grass slope — and doesn’t pretend to.
MY1016Z import-direct at $293 nominal ($320–380 delivered). Same job, but 10-in pneumatic wheels make it genuinely better on grass and gravel.
The wheelchair donor ($219 free / $319 at $100) is the step that buys curbs, mud, slopes, and 2-hour runtime. Only attempt DIY tracks ($414) if no chair can be found — and only if you’re comfortable with belt tensioning and alignment.
Drive it with the phone or a FlySky ($0–50) first. Add UWB follow later for ~$50. Autonomy is a bolt-on, not a foundation.
• MY1016Z US-retail price: only direct/Alibaba prices ($38–49) and EU/CA retail (€86 / $119 CAD) were found; no citable Amazon.com USD price on 11 Nov 2026. Stack B’s sub-$300 claim depends on the import price.
• Estimated items (24 V chargers, ST-Link, chain/sprockets, tote, buck/wiring) total $50–90 per stack — plausible ranges, not quotes.
• Power Wheels / drill / wiper torque and mower/wagon donor prices: no citable US listings found; those rows are reasoned from vehicle/tool ratings and labelled as such.
• Shipping & tax excluded throughout. US shipping on motors + battery is typically $15–40 — the gap between Stack B’s $293 nominal and $320–380 realistic.
• No live checkout was performed; prices are from the search index and pages opened 11 Nov 2026. Re-check the MY1016Z, BTS7960, and battery listings before buying.
• Printed-part strength is engineering judgement (layer adhesion / creep), not a tested datasheet for your specific resin/FDM materials.