Beyond the Robot Dog for Stair Haulage: Tracked Carriers · Humanoids · Stair-Climbing AMRs · Goods Lifts — A Horizontal Review (Draft, To Verify)

⚠️ This is a draft to verify. Survey date: June 2026. Except for the Unitree B2-W (cross-verified against both official site versions), most specs are vendor claims or media/reseller relays, flagged item by item; before settling on a selection, verify step height and slope under load with the vendor. This piece is the sibling of Robot-dog haulage selection — that one is about "which robot dog to pick," this one is about "what else exists beyond the robot dog, and when it beats the robot dog."

🌐 What this piece answers

Otsuka Shokai currently uses AGV/AMR to haul goods inside the plant; the pain point is that they can't climb stairs/ramps. Robot dogs can cover that blind spot, but their walking payload is only ~40kg (see the sibling piece). The boss's question is direct: is there another robot that can climb stairs/ramps and carry load, and is more usable than a robot dog?

This piece sorts the candidates into five categories, giving real products + specs + sources for each, and lands on "when do they beat the robot dog, and when is the robot dog still the more cost-effective choice."

🏆 Core conclusions first

1. A truly "autonomous + climbs real stairs + carries load" mature product barely exists. What you can buy today, and that really does carry loads up stairs, is essentially a "robot dog on wheels" (wheeled-legged quadruped) — represented by the Unitree B2-W and Pudu D5. These aren't a different path; they're an upgraded robot dog.
2. Everything that crushes the robot dog on payload (200–420kg) is a tracked / electric stair-climbing cart — but all need a human operator, so it's "ease the strain," not "remove the person."
3. No humanoid today can carry loads up stairs on a production line — they all do flat-floor work; stair-climbing is only a lab/marketing demo.
4. For a fixed inter-floor route with real stairs, the most rational answer is often not any robot at all, but a goods lift / VRC vertical conveyor, or "AMR on the flat + elevator/goods-lift linkage" — cheaper, more reliable, and an order or two of magnitude higher in throughput.

🛞 Category 1: Wheeled-legged platforms — the only buyable "climb stairs + carry load + autonomous" category

This category is the robot dog with wheels added, and the only realistic hardware answer to the "better suited than a robot dog" question.

Key point: the B2-W is currently the only "order it today and it really carries load up stairs" first choice — same ~40kg class as the robot dog, but with flat-ground 15km/h cruise + longer range. The Pudu D5 is the domestic Chinese equivalent with an explicit step-height spec (25cm) — essentially "a robot dog that climbs stairs." DEEP Lynx has the strongest ingress protection but only 15kg continuous payload — it carries less than a robot dog, positioned for inspection/light loads. LEVA carries 85kg but is only a research prototype, not buyable. Beyond the B2-W, most specs are reseller/media relays — before ordering, verify "step height under load" with the vendor.

🔧 Category 2: Tracked / electric stair-climbing carts — payload crushes, but needs a human operator

These are not autonomous robots; they are "power-assisted stair-climbing trolleys." Pros: big payload, mature, cheap. Con: each trip needs a person to push/remote-control.

Key point: payload of 200–420kg completely crushes the robot dog, but each trip needs a person — it's labor-saving, not labor-removing. The only truly autonomous option is a tracked "bare chassis" (the TK1000 class), but turning that into a haulage robot is a bigger engineering effort than just buying a robot dog. As of 2026.6, there is no off-the-shelf "high-payload + fully autonomous + tracked stair-climbing haulage robot." Fit: large-payload, occasional, human-operation-acceptable stair haulage.

🤖 Category 3: Humanoids — small payload, and not one can carry loads up stairs on a line

Be sure to write this into the proposal honestly: as of mid-2026, no humanoid is a production-grade "load-carrying stair haulage" solution. Every deployed humanoid logistics project (Digit/GXO, Figure/BMW, Apollo/Mercedes) hauls totes on flat ground. The one with the most future potential is Atlas (30kg sustained + the strongest motion ability), but it's still an early pilot. Most specs include vendor claims/second-hand relays, flagged item by item.

📦 Category 4: Stair-climbing / step-crossing AMRs — most "call the elevator" to bypass stairs

• Autonomous AMRs that can climb real stairs are essentially the wheeled-legged quadrupeds above (RIVR, Pudu D5).
• Star-wheel / cluster-wheel stair-climbing cargo AMRs = academic/patent, no mature product (per-step impact, wear, and the requirement that wheel diameter exceed step height make them bulky).
• Curb-delivery robots (e.g. Starship) can only handle curbs and single steps, not stairs (payload ~10kg).
• Mature indoor AMRs all "use the elevator, don't climb stairs" — as the honest alternative for multi-floor plants:
  • Relay2 (Relay Robotics), payload 45kg, integrates with Otis/KONE/Mitsubishi elevators — production-grade maturity¹⁵.
  • Pudu T300 / Keenon / Bear likewise — flat floor + elevator linkage, in production.

🏗️ Category 5: Non-robot / hybrid solutions — the real optimum for fixed inter-floor routes

The key honest conclusion: for a fixed inter-floor real-stairs route, a goods lift/VRC or conveyor is almost always cheaper, more reliable, and an order or two of magnitude higher in throughput than any legged/tracked robot. Stair-climbing robots / robot dogs only make sense in the narrow scenario of "occasional, low-payload, can't install fixed equipment" — and even then, one worker + a few-thousand-dollar electric stair-climbing cart is usually still more cost-effective. Stating this proactively when pitching Otsuka Shokai actually looks professional and builds trust.

📌 Flat-ground high-throughput is a separate line (not expanded here): for high-density/batch haulage within the same floor, the mother-child cart / tugger-train (a latent-lift mother cart towing multiple child carts) is a more mature high-throughput option than a single AGV/AMR — but it only runs on flat ground, doesn't climb stairs, so it belongs to "boosting AGV/AMR flat-ground efficiency," not "covering the stair-climbing blind spot." That line is pursued in a separate survey; this piece focuses on the stairs/ramp break point, so it isn't expanded here.

📊 Master comparison table

🎯 Ranking advice: when it beats the robot dog, when the robot dog still wins

Slot it in by scenario:

1. Fixed inter-floor route, real stairs, fairly large load → goods lift/VRC or AGV+elevator linkage (beats any robot outright).
2. Full-auto cross-floor, building has/can add an elevator → AMR + elevator linkage (e.g. Relay2); don't buy a stair-climber.
3. Must walk real stairs, must be autonomous, load ≤40kg → wheeled-legged (B2-W first choice / Pudu D5), clearly beating a plain robot dog.
4. Large load (100–420kg), occasional, human operation acceptable → tracked/electric stair-climbing cart (XSTO, SANO), the best value.
5. Continuous box flow up stairs → spiral/inclined-belt conveyor.
6. Small rise / half-floor → build a ramp for a plain AMR (note the ~5% slope ceiling, lower for heavy loads).

Scenarios where the robot dog still wins: rugged/unstructured ground, narrow passages, the need for "one machine, many uses" (mixed inspection + haulage + step-crossing), and tight budgets with small loads and variable routes — here a flexible robot dog beats wheeled-legged/tracked/humanoid/goods-lift options.

One line for the boss: don't bet the Otsuka Shokai pitch on the robot dog alone. First ask whether the route is "fixed inter-floor" or "variable," and how heavy the load is — for fixed large loads, recommend a goods lift/VRC (looks professional, no over-selling); for variable mid/small loads, go robot dog/wheeled-legged (playing to its flexibility). Making "we help you pick the right solution, not just sell you a dog" the differentiator is more likely to win the partnership than hard-pushing a robot dog.

⚠️ For you to verify / supplement

1. Real routes & goods weight: are Otsuka Shokai's cross-floor routes "a fixed few" or "variable"? How heavy is each item? These two points directly decide which row of the table above to pick.
2. Spec verification: beyond the B2-W, the specs for Pudu D5 / DEEP Lynx / tracked carts / humanoids are mostly vendor claims or media relays — before deployment, get a datasheet from the vendor and verify step height and slope under load.
3. Goods-lift/VRC feasibility: can the customer's plant retrofit a goods lift/VRC? This is often the optimum, worth assessing the civil works/cost first.
4. Whether to add models: if you want domestic Chinese tracked haulage robots, Japanese local stair-climbing cart brands, or this table turned into something filterable by "payload × route type," just tell me how to organize it.