Updated

SEER M4 landing-path deep dive: quadruped interface × can it bridge Otsuka's existing AGVs × how to partner/procure (the most likely domestic off-the-shelf path · working draft)

机器狗仙工智能SEERM4SRCVDA5050调度RCS异构AGV集成商大冢商会方案
On this page

⚠️ A working draft under continuous verification. Every link here is curl-verified 200. Three caveats: (1) the software/controller pages are first-party official; D1/D2 quadrupeds, D1-W/D2-W wheeled-legged, and the WRC2025 showing are media-sourced (zhineng518 / leaderobot) — facts are credible but phrasing follows the media. (2) M4's concrete third-party AGV compatibility list, and quadruped-specific interfaces (joint motors / whether EtherCAT/CAN-FD) are not public — anything touching concrete compatibility is flagged "verify with SEER / PoC." (3) "Quadruped + VDA5050 heterogeneous mixed dispatch" is an officially stated SEER product capability (en/media/307, media/149), but whether it covers Otsuka's fleet depends on whether Otsuka's existing AGVs support VDA5050 — the biggest variable here.

🌐 What this answers

The sibling car-dog integration case list reached one conclusion: robot-dog×AMR integration is almost entirely self-built, but SEER M4 is the one domestic off-the-shelf path where a quadruped truly enters a dispatch system (RCS). This article zooms in on SEER specifically and drills the three things the boss cares about:

  1. SEER M4 quadruped interface details — how does a quadruped actually plug in? What interfaces/protocols?
  2. Can it bridge Otsuka's existing AGVs — Otsuka has an existing wheeled AGV fleet; can M4 dispatch them all?
  3. How to partner/procure — how to buy, how to cooperate, is there a Japan presence?

Related: full-lineup haulage selection → Haulage selection; starting with GO2 → GO2 selection; the three fusion forms → Fusion; the minimal self-built architecture → Self-built middleware architecture.

🧩 1. SEER M4 quadruped interface details

① What M4 is

M4 is SEER's third-generation dispatch product, an all-in-one intelligent robot dispatch/management system — the RCS layer in customer terms. Official: "M4 智能机器人调度系统是仙工智能打造的第三代调度产品" "over 8 years of project accumulation and 1000+ deployments"1. Scale, official: "supports million-level storage locations and 100k-level points; a single system handles 100+ robots of multiple types per zone, or 300+ of a single type"2.

② How a quadruped plugs in — a "two-layer stack," not either/or

The most misread point: integrating a quadruped with SEER is bottom motion control + top dispatch access, two layers stacked.

Layer Does what With what Official/media quote Source
Bottom · motion control makes the quadruped "walk and jump" quadruped carries an SRC controller the dog "carries SEER's control system," "active disturbance-rejection RL algorithms + high-torque joints," can "walk flexibly, move omnidirectionally, jump" 13
Top · dispatch access makes the quadruped "enter the fleet, be unified-dispatched" plugs into the M4 dispatcher "the dog supports accessing the M4 AI logistics management system, unified dispatch + cooperation with the plant's other types/brands of robots" 13
  • The SRC controller is officially the robot's "super-brain" on a 100%-self-developed architecture; the latest SRC-5000 is "fully upgraded in compute," and "by end-2024 SRC controllers had adapted 300+ components"14.
  • Key capability (prerequisite for third-party mechanisms): "SEER's controller is currently the only one supporting secondary development, providing integrators a scripting dev environment + rich API; for complex mechanisms it supports motion control of up to 12 servo motors"7.

③ External interfaces / protocols (what a self-built integration actually targets)

  • M4 externally = HTTP API + low-code scripts: "complete HTTP API to query/modify the system, place orders, with callbacks," "script-extend support for third-party devices like special doors/elevators," "fully open API, extend business logic via UI low-code and Python scripts"21.
  • Industrial protocols: "M4 core supports common PLCs and Modbus, OPC, S7 industrial protocols," covering doors, elevators, lifts, weighers, visual servos2.
  • SRC controller externally: provides TCP/IP, Modbus APIs + scripting dev environment for integrators to do secondary development9; hardware interfaces include RS232/RS485/CAN/Gigabit-Ethernet/USB/DI·DO, with SRC-3000FS listing "5× Gigabit industrial Ethernet + IEEE1588v2 (PTP)"10.
  • Tooling (official software matrix): Roboshop Pro (one-stop low-code implementation tool for integrators), Robokit (robot OS), RDS, MWMS, Meta-V, RoboView8.

④ Robot dogs SEER has integrated

  • D1 / D2 are SEER-ecosystem quadrupeds (media-confirmed; ⚠️ correction: the official wording is "together with partners" and "equipped with the SEER control system" — partners build the dog, SEER supplies the brain, not SEER-made complete units; the hardware maker is unnamed, see SEER marsupial feasibility; unrelated to the Unitree D1 6-axis arm): D1 = navigation version (all-terrain logistics), D2 = patrol version (all-weather plant inspection), both with 3D-SLAM autonomy13.
  • Wheeled-legged D1-W / D2-W (CeMAT): D1-W "navigation wheeled-legged dog, optional point-foot/wheel-foot," "whole-unit IP54," "0–40°C wide-temp"; D2-W "patrol, with thermal/pickup/strobe"14.
  • WRC2025: "two SEER-controller quadrupeds shown, one with an inspection payload"14. ⚠️ No public source names the specific third-party quadruped brands at WRC2025 (media only says "quadruped") — to tell the client "SEER has integrated Unitree etc." you must verify with SEER directly.

🔗 2. Can it bridge Otsuka's existing AGVs

Conclusion: yes — via VDA5050, and contingent on whether Otsuka's existing AGVs support VDA5050.

① Heterogeneous / third-party unified dispatch is a headline M4 selling point

M4 "adapts SEER's full robot lineup and some third-party models," "multi-zone (e.g. floors), multi-type"12. ⚠️ Official says only "some third-party models," with no public brand/model list — verify with SEER or by PoC.

② VDA5050 — the standard path to bridge other AGVs (officially supported)

  • "SEER Robotics SRC series core controllers have been fully adapted to the VDA5050 protocol"; SRC-equipped robots "can seamlessly access the same scheduling system with other brand models"5.
  • The strongest line (directly hitting Otsuka's case): "Whether robots powered by SEER's control system or third-party robots compatible with VDA5050, all can collaborate seamlessly within a single scheduling system," explicitly covering "lifting robots, autonomous forklifts, tote-handling robots, as well as humanoid robots and quadruped robots," stressing "avoid vendor lock-in" and "new devices integrate without restructuring the architecture (plug-and-play)"6.
  • VDA5050 substrate: MQTT + JSON, released 2021, 2.0 in 20225.

③ Grafting SEER's quadruped dispatch onto Otsuka's fleet — realistic paths (by preference)

Path How Pros Hard gate / risk
① VDA5050 bridge (most realistic · keep the fleet) M4 as unified RCS: SEER-ecosystem quadrupeds (D1/D2 on native SRC) + Otsuka's existing AGVs (via VDA5050) mixed dispatch officially supported, keeps the investment, plug-and-play Prereq: Otsuka's AGV vendors provide a VDA5050 interface — biggest variable, verify per unit
② Buy SRC controllers only, self-build the dog SRC for motion control + Roboshop Pro secondary dev; dispatch ties into Otsuka's own host (M4 HTTP API) no full SEER swap, fits "in-house software × off-the-shelf hardware" upper-layer dispatch integration is on us; quadruped-specific interface to confirm
③ Full SEER swap M4 + SEER's full robot lineup smoothest dispatch abandons fleet investment, conflicts with "keep existing AGVs," not recommended
④ Pure self-build (reference only) use SEER as benchmark, build a VDA5050 RCS fully autonomous highest cost/longest timeline, fallback only if talks fall through (see Self-built middleware architecture)

🤝 3. Partnership / procurement

① Japan presence / agent — none found locally (a key risk)

  • The only overseas subsidiary is in Germany: "established an overseas subsidiary in Germany, along with domestic offices in Hong Kong, Wuxi, Shenzhen, Chengdu," serving 65+ countries but no Japan agent/distributor listed11.
  • There is a Japanese site (signals Japan intent): seer-robotics.ai/jp, in Japanese, but the page has no Japan base/agent/Japanese support line — inquiries still route to HQ email12.

② Procurement / contact

  • General email contact@seer-robotics.ai; CEO ceo@seer-robotics.ai1112.
  • SRC / M4 procurement: no online pricing, quote-based; SRC product pages list specs (SRC-800/2000/2000-F(S)/3000FS/5000)94.

③ Integrator-facing model — fits Otsuka precisely

  • Official "product portfolio strategy for integrators": SRC + Robokit + Roboshop Pro low-code lets integrators "make secondary creations and tailored implementation solutions," "a STEM graduate learns to deploy SEER software in a week"8.
  • So as a systems integrator, Otsuka can take the "buy SRC controllers / M4 license + do upper-layer business integration in-house" OEM/partner form — SEER sells exactly "zero barrier for integrators to build robots," a natural fit for "in-house software × off-the-shelf hardware."

🎯 4. Honest conclusion for the boss

  1. SEER is the one domestic off-the-shelf path that has made "quadruped + VDA5050 heterogeneous mixed dispatch" an officially stated product capability (M4 + SRC + Roboshop Pro integrator toolchain) — far more realistic than writing a VDA5050 bridge from scratch; referenceable or even partnerable.
  2. Most realistic form: M4 as unified RCS, SEER-ecosystem D1/D2 (partner hardware) on native SRC + Otsuka's existing AGVs via VDA5050 — fitting "in-house software × off-the-shelf hardware + keep the existing fleet."
  3. Two hard gates decide it, verify first:
    • Whether Otsuka's existing AGVs support VDA5050 — the biggest variable for Path ①; verify each fleet vendor. If not, it degrades to Path ② (buy SRC, self-build).
    • SEER has no Japan agent — deal direct with Shanghai HQ / Germany; Japan localization/lead-time/language capability unknown — ask early.
  4. Clear fallback: even without a SEER deal, use it as a benchmark and self-build — the minimal architecture is in Self-built middleware architecture.

⚠️ For you to verify / to add

  1. Otsuka's existing AGV vendors + VDA5050 support — the first gate of the whole path; give me the fleet vendors/models and I'll check VDA5050 status one by one.
  2. The third-party quadruped brands integrated at WRC2025 / real projects — unnamed officially; want me to draft an inquiry email to SEER (contact@seer-robotics.ai) covering the third-party compatibility list, quadruped interface details, and Japan support?
  3. SRC quadruped-specific specs (joint motor models, EtherCAT/CAN-FD, the quadruped SRC model) — only generic AMR figures are public; SEER must provide.
  4. Procurement/partnership form — want me to draft a separate inquiry/cooperation checklist for SEER for each of Paths ① (VDA5050 bridge) and ② (buy SRC, self-build), mirroring the GO2 EDU procurement checklist?