Updated

Candidate Entry Scenarios · Feasibility Quick-Read: Running the Numbers on 3 Scenarios + Priority (Draft, To Verify)

切入场景可行性巡检交互机器人数据采集选型任务匹配
On this page

⚠️ This is a draft to verify. The form-factor/capability judgments are backed by verbatim official lines (see the first two pieces; the end-of-sentence superscripts jump to the source); but costs are rough estimates, and channel & customer fit are inferred from the assumption "Otsuka Shokai = a general trading-company channel serving Japanese enterprises" — all flagged "to verify." Please have the boss correct them with real quotes and real customer data before deciding.

⚠️ Note: DEEP Robotics' official site deeprobotics.cn had its TLS certificate expire on 2026-06-25 (yesterday); the related links in this piece are temporarily unreachable and will recover once the cert is renewed — DEEP figures are treated as "to verify" meanwhile.

🌐 What this piece answers + the evaluation method

The first two pieces (Robot-dog capability map, Other embodied-AI map) nailed down "who can do what." This one brings it down to "which one should our company build first": three candidate scenarios, each run through the same 5-item quick-read, then a priority ranking and a recommended first pick — producing a decision the boss can act on.

The 5 quick-read dimensions (applied uniformly to each scenario): ① which form factor (per the form-factor-fit conclusions of the first two pieces) ② rough cost (base + add-ons, drawing prices from the capability research, flagged to verify) ③ the unique value of our own software/AI (what our company can add that others can't) ④ Otsuka Shokai channel/customer fit (will existing customers pay for it) ⑤ ramp-up difficulty and risk.

Three candidate scenarios: (a) facility inspection = robot dog + in-house inspection software; (b) expo / reception interaction robot; (c) mobile data-collection platform.

🔍 Scenario A: Facility inspection (robot dog + in-house inspection software)

Routine inspection of factories/substations/data centers/utility tunnels — reading gauges, taking temperatures, spotting abnormal sounds, checking for leaks/drips/spills — replacing people in hazardous or repetitive on-site work.

Quick-read item Conclusion (to verify)
① Which form factor Quadruped robot dog. It lands in all of its green zones: ① mobility/obstacle-crossing + ④ perception + ⑤ autonomy (incl. GPS-denied) + ⑪ outdoor/explosion-proof certification. The official positioning is literally "navigating complex, multi-floor plants made for humans"5
② Rough cost Entry validation uses a Go2-class consumer tier (low)1; formal industrial deployment uses Spot3/ X304/ ANYmal5/ B22, with a single industrial-tier base around ¥several-hundred-thousand to ¥1.5M+ (reseller/official figures vary widely, to verify, see Robot-dog product research); add-ons: thermal-imaging/gas/audio-pickup sensors + charging dock
③ Unique in-house software/AI value Highest. The base is bought; all the differentiation is in software: inspection route orchestration, gauge-reading OCR, thermal/vibration anomaly detection, reporting and alerting, integration with the customer's MES/SCADA. Delivered via ⑧ open SDK/ROS (Lite3 "motion control SDK and APIs… with sample code", Vision 60 "C/C++, ROS, ROS2")
④ Channel/customer fit High. Japanese manufacturing/energy/infrastructure have large installed facility bases + labor shortages + hard safety-compliance needs; the general trading-company channel buys into the "replace human inspection, cut cost, stay compliant" narrative (fit to verify against real customers)
⑤ Ramp-up difficulty and risk Medium. The base is mature; risk is concentrated in software integration and on-site adaptation (networking/explosion-proof zones/customer-system integration); not frontier hardware R&D, so it's controllable

Quick verdict

A is the standard answer for the robot dog's home turf: all-green form fit, highest in-house software value, smoothest channel narrative, controllable technical risk. The only variables are the industrial base cost and the depth of customer-system integration.

🎪 Scenario B: Expo / reception interaction robot

Greeting, guiding, narrating, and drawing crowds in showrooms/expos/lobbies — friendly interaction and performance aimed at people.

Quick-read item Conclusion (to verify)
① Which form factor Not a robot dog. This lands in the quadruped's ⑥ human-interaction red zone (a low, squat form is not friendly). The most cost-effective form is a wheeled service robot with a screen (⑥ interaction + ⑦ runtime + ⑨ maturity + ⑩ cost all green), like the BellaBot class "3D Omnidirectional Obstacle Avoidance" + 24/7 hot-swap7; only go humanoid when budget is ample and you want talking-point value (G1 around US $13.5K8, but weak on runtime/maturity)
② Rough cost A wheeled base with a screen is relatively low (tens of thousands to ~¥100K+ RMB class, to verify); humanoid G1 starts around US $13.5K8; a robot dog forced into this role needs added screen/voice/upper-body modification, with poor value for money
③ Unique in-house software/AI value High and on-trend: multimodal dialogue (LLM), face/foot-traffic analytics, multilingual narration, content orchestration and data capture. Voice interaction/narration AI is where our company can strongly differentiate
④ Channel/customer fit Medium. Expos/retail/reception are project-based — high on talking-point value but unstable in unit price and repeat purchase; the general trading company can do it, but it's opportunistic, not a stable revenue base
⑤ Ramp-up difficulty and risk Medium. The base is mature; risk is in polishing the interaction experience and on-site reliability (recognition in noisy environments, long-run operation); picking the wrong form (using a dog for reception) is the main pitfall

Quick verdict

B has high AI value but strays from the robot dog's home turf: the form must switch to wheeled-with-screen, off the boss's main line of "use the robot dog as the platform first." It suits an opportunistic pilot / brand exposure, not the main thrust.

📦 Scenario C: Mobile data-collection platform

Let a walking base carry multiple sensors (LiDAR/thermal/gas/RF/camera) and autonomously collect data in plants, construction sites, agriculture/forestry, post-disaster environments, etc., feeding it back for 3D reconstruction/inspection/AI training.

Quick-read item Conclusion (to verify)
① Which form factor Quadruped robot dog (same base as A). It's essentially a "walking sensor base," right in the sweet spot of ② platform-grade payload + ④ integrated perception: B2 ">40kg" payload2, Go2 "Standard Ultra-wide 4D LiDAR"1; it covers the blind spots of drones in rough/indoor terrain. For complex off-road you can stack the wheel-legged B2-W9
② Rough cost Base same as A (to verify); the increment is mainly in the sensor payload (high-precision LiDAR/multispectral can cost more than the base) + the data pipeline. It can share a base with A, so marginal cost is low
③ Unique in-house software/AI value High: collection-task orchestration + synchronized localization + point-cloud/imagery AI processing + turning data into an asset. It shares autonomous navigation and the software stack with A, so reuse is high
④ Channel/customer fit Medium–high. Surveying/construction/energy/research have real data-collection needs, but the customer base is more scattered than "inspection cost-cutting" and must be sold industry by industry
⑤ Ramp-up difficulty and risk Medium. The base shares A's risks and is controllable; the hard parts are high-value sensor selection + data-processing delivery quality; customers demand "usable data"

Quick verdict

C is A's neighbor, sharing the base and software stack: low technical risk, reusing A's investment. The customer base is more scattered than A's, so it suits being a second-step expansion after A, not a standalone first launch.

Ranking the three scenarios on a composite of "form-factor fit × in-house software value × channel fit × controllable risk" (scoring is analysis, to verify):

Rank Scenario Form In-house AI value Channel fit Risk Composite
1 A Facility inspection Quadruped 🟢 High 🟢 High 🟢 Controllable First entry pick
2 C Data collection Quadruped/wheel-legged 🟢 High 🟡 Medium-high 🟢 Controllable Second step after A (shared base)
3 B Interaction robot Wheeled w/ screen 🟢 High 🟡 Medium 🟡 Medium Opportunistic pilot

💡 Recommendation: lead with A (facility inspection) as the first entry, C as the second step, B as an opportunistic pilot

  1. Launch with A — it's the only all-green scenario in the robot dog's capability profile: a perfect form-factor match, the largest in-house software differentiation space, the smoothest narrative for Japanese manufacturing/energy/infrastructure channels, and technical risk concentrated in controllable software integration. This is exactly what confirms the boss's instinct that "the robot dog is low-cost, use it as the platform first."
  2. Once A is working, follow up with C — the same base + autonomous navigation + software stack, expanding "inspection" into a "collection + inspection" data platform, with low marginal cost and high reuse.
  3. B as an opportunistic pilot — high AI value and good brand exposure, but it requires switching to a wheeled-with-screen form and strays from the robot-dog main line; recommend taking it on project by project, not committing the main force.

One-line decision recommendation: First use a robot dog + in-house inspection software to enter "facility inspection" (Scenario A), proving out the in-house software value and the trading-company channel; then expand to data collection (C) on the same base; switch the interaction robot (B) to wheeled-with-screen and take it on opportunistically. Please correct all cost/channel figures with real quotes and customer data.

⚠️ For you to verify / to be supplemented

  1. Cost: the real landed price of an industrial base (incl. Japan procurement/warranty/giteki), plus add-on sensors and software person-months — see Robot-dog product research, still to verify.
  2. Channel/customer: "which scenario types Otsuka Shokai's existing customers will pay for" is this piece's biggest assumption variable; please use a real customer list/industries to correct the priority.
  3. Whether to detail a given scenario: for the first pick, Scenario A, we can produce a landing plan of "base selection + add-on BOM + software modules + a pricing framework + a pilot roadmap" — just let me know.