What is an autonomous mobile robot (AMR)?

What AMR means

Autonomous mobile robot (AMR) is the category of robots that navigate dynamic environments without fixed paths. Per DEPLOY's framework on physical AI, AMRs are one of the four subcategories DEPLOY tracks under the physical AI umbrella. The defining technical bet is onboard perception + planning: the robot decides its own route through the environment using sensors (lidar, cameras, IMU) rather than following predefined markers or paths.

Two structural distinctions matter for the AMR category:

1. AMR vs AGV (automated guided vehicle). AGVs follow fixed routes marked by floor tape, magnetic strips, or QR codes. AMRs navigate dynamic environments via onboard perception. The AGV category is older, more rigid, and operates at lower per-unit cost; AMR is the newer category with higher capability and broader use-case applicability.
2. Free-roaming AMR vs grid-AS/RS (automated storage and retrieval system). Grid-AS/RS platforms (AutoStore, Ocado, Symbotic at scale) operate on predefined grid infrastructure where robots traverse rails or rack structures. Free-roaming AMRs (Locus, MiR, Geek+, Berkshire Grey, Fetch) navigate open warehouse floors via onboard sensors. The two architectures have structurally different capital-deployment profiles + operational envelopes.

Many commercial deployments combine both architectures: grid infrastructure for high-density storage + free-roaming AMRs for pick-pack-ship workflows.

---

The AMR cohort DEPLOY tracks

Per DEPLOY's registry, the major AMR cohort includes:

• Symbotic: grid-AS/RS warehouse automation at meaningful scale. Walmart relationship anchors deployment depth (42 distribution centers; $22.7B backlog). See what is Symbotic for entity deep-dive.
• Locus Robotics: free-roaming AMR for pick-pack-ship workflows in third-party logistics warehouses.
• Boston Dynamics Stretch: purpose-built case-handling AMR for warehouse trailer-unloading and palletizing tasks. Distinct product line from Boston Dynamics's Atlas humanoid program.
• Zebra/Fetch Robotics: free-roaming AMR cohort acquired by Zebra Technologies; warehouse + manufacturing applications.
• Geek+: Chinese AMR platform with broad enterprise deployment.
• Berkshire Grey: AMR + AI-perception integration for retail + parcel handling.
• AutoStore: grid-AS/RS pioneer; storage cube architecture with picking robots above the grid.
• Ocado: grid-AS/RS at automated grocery fulfillment scale; UK-based with international licensing.
• Amazon Robotics Fleet: captive (internal-not-customer-sale) AMR cohort operating across Amazon fulfillment centers globally. Boston Dynamics Stretch operates external customer deployments; Amazon Robotics operates Amazon-internal scale.
• MiR (Mobile Industrial Robots): Danish AMR platform; manufacturing + intralogistics applications.

---

Captive vs customer-sale deployment distinction

Per DEPLOY's maker-facility rule, captive deployments (where the operator and the maker are the same entity) classify differently from arms-length customer deployments. Amazon Robotics Fleet is the canonical captive AMR cohort: Amazon-internal operational scale across Amazon fulfillment centers globally. The captive scale is meaningful as engineering credibility but does not constitute customer-sale verification in the same way Locus, Symbotic, MiR, or Stretch deployments at external customers do.

The distinction matters because captive deployments do not face the customer-procurement verification gate (negotiation, integration, service contract, multi-vendor competition) that external deployments face. The framework treats Amazon Robotics's scale as engineering-credibility surface rather than customer-validated commercial deployment.

---

Where AMRs fit in physical AI category

Per DEPLOY's physical AI category framework, AMRs are one of the four subcategories DEPLOY tracks alongside autonomous vehicles + humanoid robots + AI-augmented industrial automation. The AMR subcategory has distinct verification framework dimensions:

• Operational design domain: warehouse + manufacturing + distribution center contexts; structured industrial environments with controllable variation.
• Regulatory framework: OSHA workplace safety primary; not the NHTSA SGO or CPUC frameworks that apply to AVs.
• Labor-market interface: warehouse worker augmentation + replacement at structured-task pilot scale.
• Deployment cadence: faster commercial deployment cadence than humanoids; slower per-task generalization than AVs because tasks are more specialized.

DEPLOY's four canonical frameworks (availability + capability + safety + value chain) apply to AMRs with subcategory-specific worked examples.

---

Verification framework applied across AMR cohort

Per DEPLOY's verified-vs-claimed framework:

• Verified deployment scope: Symbotic at 42 Walmart distribution centers + $22.7B backlog; Locus at multi-warehouse 3PL deployment; Stretch at external customer trailer-unloading operations; Amazon Robotics at Amazon-internal scale (engineering-credibility captive).
• Per-deployment throughput: varies by operator + platform; published metrics inconsistent across cohort.
• Pricing: enterprise-contract bound for customer-sale AMRs; not publicly disclosed at consumer-evaluation depth.
• Captive vs customer-sale distinction: Amazon Robotics Fleet operates at captive scale; per maker-facility rule, classifies as engineering-credibility rather than commercial deployment.

---

Bottom line

AMRs are the category of robots that navigate dynamic environments via onboard perception + planning rather than fixed paths. Two architectural variants: free-roaming AMRs (Locus, MiR, Geek+, Berkshire Grey, Fetch, Stretch) navigate via sensors; grid-AS/RS platforms (AutoStore, Ocado, Symbotic-at-scale) operate on predefined grid infrastructure. The cohort spans customer-sale deployments (Symbotic, Locus, Stretch external) and captive deployments (Amazon Robotics Fleet) with distinct verification postures.

For broader physical AI category context, see what is physical AI. For Symbotic specifically, see what is Symbotic. For the cohort's relationship to DEPLOY's canonical frameworks, see can humanoid robots replace workers (which covers workforce-replacement framework applicable to AMR deployments) and how DEPLOY verifies deployment status. For methodology canonical references applicable to AMR category: the 4-way autonomy-boundary taxonomy (warehouse autonomy-boundary mapping) + the 9-tier source-quality rubric.