Patent 06 / Edge-Node Resource Distribution
01 / 11 US10958583B2
↑↓  navigate  ·  all patents →
Siten Sanghvi  ·  Granted March 23, 2021

The Network
Moves the
Display.

A consensus protocol that lets edge-nodes sense, agree, and act — autonomously repositioning product displays and rerouting cloud resources without asking anyone's permission.

US10958583B2Patent number
2019-06-26Filed
1 yr 9 moTime to grant
13 Claims / 2 independentClaim count
18 CitationsForward citations
SCROLL TO EXPLORE
Visual patent explainer
01 / The Problem

Static systems in a dynamic world.

Cloud infrastructure and physical retail layouts share the same flaw: they're configured once and left. Neither responds to what's actually happening in the space, right now.

A display placed at opening is still there at closing — regardless of whether anyone is engaging with it. Cloud resources allocated to a branch stay fixed regardless of whether that branch is busy or empty.

Three failure modes
01
Static display placementProduct displays positioned manually, held fixed regardless of customer flow or engagement patterns throughout the day.
LAYOUT
02
Cloud-blind allocationComputing resources distributed to locations by schedule or plan, not by real-time demand or measured local traffic.
COMPUTE
03
No local intelligenceEvery adaptive decision requires a round-trip to the cloud — latency and dependency that defeats the purpose of local responsiveness.
LATENCY
02 / The Invention

Consensus at the edge.

Three edge-nodes, each sensing a different signal, run a consensus protocol together. No single node decides. No cloud required. The result: the display moves to the right position and cloud compute flows to where it's needed most.

01
Distributed Sensing

Three nodes independently measure total footfall, display-attracted traffic, and direct customer engagement — three orthogonal signals from the same physical space.

Richer signal — measuring the same environment from different angles produces a more robust interest estimate than any single sensor could achieve alone.
02
Consensus Protocol

The edge-nodes collectively determine the current level of interest and compute the position that would maximize it — with no central authority.

BFT / proof-of-work / proof-of-stake — the protocol is implementation-agnostic; any Byzantine-fault-tolerant consensus mechanism qualifies under the claims.
03
Autonomous Actuation

The embedded node receives the consensus output and physically moves the display — then re-senses, closing the feedback loop without human intervention.

Sense → agree → act → re-sense — the complete loop runs at the edge, latency-free, with no cloud dependency at actuation time.
03 / Architecture

Three nodes. One protocol. One outcome.

The system claim defines a three-node edge-computing system. Each node captures a distinct signal. All three run a shared consensus protocol whose output is both physical — the display moves — and computational — cloud resources are rerouted.

Edge-node system — Claim 1 structure
1
Traffic Sensor NodeSenses total volume of human customer traffic flowing through the banking center.
2
Display Traffic NodeSenses the volume of customer traffic specifically attracted by the moveable product display — not ambient footfall.
3
Embedded Display NodePhysically embedded inside the display. Senses direct customer engagement — dwell time, interaction, proximity.
↓  consensus protocol  ↓
Consensus Output Determines interest level → calculates optimal new position → issues movement instruction to the display
04 / Consensus Protocol

Agreement without a central authority.

The three edge-nodes run a Byzantine-fault-tolerant consensus algorithm. No single node can move the display unilaterally — agreement requires the collective data from all three, which eliminates single points of failure.

Five-step consensus execution
1
SenseEach node independently captures its signal: overall footfall, display-attracted traffic, or direct engagement data.
2
ShareThe three nodes exchange readings across the local edge network — no cloud roundtrip required.
3
AgreeConsensus protocol runs (BFT, proof-of-work, or proof-of-stake). Nodes converge on the "default level of interest" at the current position.
4
CalculateThe agreed model computes a second position within the banking center projected to increase the measured interest level.
5
ActNode 3 — embedded in the display — receives the movement instruction and physically repositions the display.
05 / Sensor Architecture

Two node types. One feedback loop.

The system distinguishes sensor nodes (read the environment) from actuator nodes (execute physical changes). The embedded display node is both — it senses engagement and receives movement commands from the consensus output.

Node type taxonomy
01
Sensor Node — EnvironmentPlaced in the banking center. Measures total human traffic volume. Passive — reads, doesn't act.
02
Sensor Node — DisplayPositioned to observe the product display. Measures traffic specifically attracted by the display, independent of ambient footfall.
03
Actuator Node — EmbeddedPhysically embedded in the display hardware. Senses engagement AND receives move instructions. The only node that causes a physical change.
04
Cloud InterfaceEdge-nodes can redirect QoS resources from a connected cloud environment — compute follows engagement, not a fixed schedule.
06 / Embedded Actuator

The display is a node.

The third edge-node isn't observing the display from outside — it's embedded inside it. The display itself is both sensor and actor, closing the feedback loop directly at the physical object without any external coordinator.

Sense → agree → act → re-sense loop
■
Embedded Node Senses EngagementMeasures dwell time, touch interaction, or proximity at the display's current position — continuous, passive monitoring.
⇄
Data Shared with Sibling NodesEngagement reading joins the shared pool with overall footfall and display-attracted traffic data from Nodes 1 and 2.
○
Consensus Runs — New Position CalculatedProtocol determines optimal position. Embedded node receives the calculated target coordinates as a move instruction.
AGREE
→
Display Moves — Loop RestartsPhysical movement executed. The embedded node immediately begins sensing engagement at the new position, feeding the next consensus round.
MOVE
07 / Cloud Resource Routing

Compute follows engagement.

The independent method claim scales the consensus protocol across multiple banking centers. Edge-node groups at each location collectively determine relative interest levels and route cloud QoS resources to where demand is highest.

Multi-location resource distribution — Claim 8
1
First Edge-Node GroupAt a target banking center. Determines level of interest in the moveable product display via local consensus protocol.
2
Second Edge-Node GroupAt the same or other banking centers. Collectively determines QoS resources currently available from the cloud environment.
3
Identify Optimal LocationCross-location consensus targets the banking center where the moveable display will generate maximum engagement uplift.
4
Route Cloud ResourcesQoS compute redirected from the cloud to the identified location. High-engagement branches get more; low-traffic ones get less — automatically.
Claim 8 is independent of Claim 1's system structure — it covers any implementation where edge-node groups collectively route cloud QoS resources based on measured interest levels.
08 / Applications

Where edge-consensus applies.

The claims define a general framework: sensor nodes, actuator nodes, consensus protocol, physical-digital action loop. The domain is wherever distributed sensing must drive autonomous physical or resource outcomes at the edge.

Express = explicitly claimed  ·  Inferred = reasonable extrapolation
Express Retail Display Optimization Auto-position product displays in retail centers based on real-time customer engagement measured by edge-node consensus.
Express Cloud QoS Distribution Route cloud computing resources dynamically across distributed locations based on measured interest levels — no human scheduling required.
Inferred Smart Retail Layouts Any commercial space where physical fixtures respond to real-time foot traffic patterns without staff intervention.
Inferred IoT Mesh Actuation Sensor-actuator networks where no single node decides — distributed consensus drives autonomous physical action across the mesh.
Inferred Edge-Native Resource Arbitration Multi-site networks where local edge groups bid for shared cloud compute based on real-time demand signals — no central allocator.
Inferred Autonomous Manufacturing Cells Factory floor systems where equipment position or compute allocation adjusts dynamically to production flow, without cloud dependency.
09 / Forward Citations

Cited by Amazon — and 7 more.

18 forward citations verified on Google Patents. Amazon Technologies dominates the forward citation graph — 11 patents citing this work, concentrated in radio-based private networks and edge QoS distribution.

Forward citations by Google Patents / checked 2026-06-16
Managing assignments of network slices Amazon Technologies, Inc. / US11252655B1 Granted 2022
Highly available data-processing network functions for radio-based networks Amazon Technologies, Inc. / US11729091B2 Granted 2023
Extending cloud-based virtual private networks to radio-based networks Amazon Technologies, Inc. / US11838273B2 Granted 2023
Demand-based allocation of ephemeral radio-based network resources Amazon Technologies, Inc. / US11895508B1 Granted 2024
System and method for improving security of personally identifiable information Truata Limited / US11263347B2 Granted 2022
Systems and methods for use in balancing network resources Ipco 2012 Limited / US20240323139A1 Published 2024
6 of 18 forward citations shown. Amazon Technologies accounts for 11 of 18. Full list available on Google Patents.
10 / Timeline

Filed June 2019.
Granted March 2021.

Priority established June 2019. Published December 2020. Granted just 3 months later — active through September 2039.

Jun 2019
Priority Filed
US16/452,862 filed
priority date established
18 mo
Dec 2020
Published
US20200412654A1
claims publicly visible
3 mo
Mar 2021
Granted
USPTO granted US10958583B2
13 claims, 2 independent
~18 years
Sep 2039
Expires
Adjusted expiration
per Google Patents
Read the granted patent View the application Browse all granted patents
End / Patent 06