Lenskart — Tote Movement Optimization

Problem

At peak load the conveyor network between the ASRS exit and GTP stations became the bottleneck — totes queued, pickers starved, and system latency bloated. The existing router used naive FIFO + static rules that ignored congestion and alternative paths.

My Contribution

• Modelled the conveyor network as a weighted directed graph — nodes are junctions/diverts, edges carry travel-time + live congestion weight.
• Implemented a Dijkstra-based shortest-path service in Java, with congestion weights updated from live conveyor telemetry.
• Added fallback routing (second-best path) when the primary route exceeded a congestion threshold, avoiding deadlocks during burst load.
• Benchmarked against production traces before pilot rollout.

Tech Stack

• Java
• Spring Boot
• Graph Algorithms
• MySQL
• RabbitMQ
• Grafana

High-Level Design

flowchart LR
  ASRS["ASRS Exit"] --> ROUT["Tote Router Service
(Java)"]
  TELE["Conveyor Telemetry"] --> ROUT
  ROUT -->|path plan| CCTRL["Conveyor Controller"]
  CCTRL --> CONV[("Conveyor Network
(graph)")]
  CONV --> GTP["GTP Stations"]
  ROUT --> METRICS["Grafana
(travel time p95)"]
                

Low-Level Design — Route Planning

flowchart TB
  A["Receive tote
(source, destination)"] --> B["Load graph snapshot
(edges + live weights)"]
  B --> C{"Primary path
available?"}
  C -- yes --> D["Dijkstra shortest path"]
  C -- no --> E["Yen's K-shortest
(k=2)"]
  D --> F["Estimate ETA + congestion"]
  E --> F
  F --> G{"ETA > threshold?"}
  G -- yes --> H["Fallback path"]
  G -- no --> I["Commit path"]
  H --> I
  I --> J["Publish route.assigned → controller"]
                

Impact

• −15% tote travel time.
• −200+ ms system latency per tote.
• +12% warehouse throughput during pilot phase.