UPS Healthcare — Movu Robotic Shuttle Integration

Problem

UPS Healthcare deployed Movu robotic shuttles on top of an existing ASRS. The two stacks spoke different protocols and kept their own state, so operators had no single view of shuttle health. Incidents (stalled shuttles, failed handoffs, lift timeouts) took hours to triage and blocked pharma-grade throughput SLAs.

My Contribution

• Engineered the integration layer connecting Movu shuttle firmware APIs with the ASRS control plane — unified orchestration across both stacks.
• Designed a real-time telemetry pipeline on MongoDB ingesting shuttle & lift events at 99.9% capture rate.
• Built a shuttle-health dashboard exposing connectivity, battery, queue depth and fault states — cut downtime diagnosis by 70%.
• Hardened the failure path: idempotent consumers, DLQs, and replay from the event store.

Tech Stack

• Java
• Spring Boot
• RabbitMQ
• MongoDB
• REST
• Grafana
• Docker
• Kubernetes

High-Level Design (HLD)

flowchart LR
  SH["Movu Shuttles
(50+)"] -->|REST / TCP| ADAPT["Movu Adapter
Service"]
  LF["Automated Lifts"] --> ADAPT
  ADAPT --> Q[(RabbitMQ)]
  Q --> ING["Telemetry Ingestor"]
  ING --> M[(MongoDB
time-series)]
  Q --> ASRS["ASRS Core
(Spring Boot)"]
  ASRS <--> DB[(MySQL)]
  M --> DASH["Health Dashboard
(Grafana)"]
  ASRS --> UI["Operator UI"]
                

Low-Level Design — Telemetry Path

sequenceDiagram
  autonumber
  participant S as Shuttle
  participant A as Movu Adapter
  participant Q as RabbitMQ (topic: shuttle.telemetry)
  participant I as Ingestor
  participant M as MongoDB
  participant G as Grafana

  S->>A: heartbeat + status (1Hz)
  A->>Q: publish telemetry event
  Q->>I: consume (idempotent)
  I->>M: upsert shuttle state (doc)
  I-->>Q: ack
  G->>M: query (last 5m window)
  G-->>Operator: live shuttle health panel

  Note over Q,M: DLQ + replay path for transient failures
                

Impact

• 99.9% telemetry capture rate across 50+ shuttles.
• Downtime diagnosis time −70%.
• Zero-data-loss event replay during RabbitMQ broker failovers.