Contact

Type to explore Drawk studio productions

Discover cross-platform games, mobile apps, architecture articles, and interactive GLSL labs.

← Projects Drove Social Platform

Drove Social Platform

Drove · · Software Engineer / Web + Interactive Developer · 2012-11-01 – 2016-06-01

Drove Social Platform
ENTERPRISE CASE STUDY · TECHNICAL DEEP DIVE

Drove: Low-Latency Geospatial Driver Matching and Telemetry Platform

Engineering real-time geospatial driver dispatch, low-latency telemetry streaming, distributed state management, and high-concurrency routing orchestration.

CLIENT / ORGANIZATION Drove
ROLE & ACCOUNTABILITY Software Engineer / Web + Interactive Developer
TIMELINE 2012-11-01 – 2016-06-01
DEPLOYED PLATFORMS Web

Executive Summary

Drove engineered a real-time ride coordination and spatial dispatch network connecting thousands of active drivers with passenger demand. We built an event-driven distributed system featuring sub-second geospatial index lookups, continuous vehicle telemetry streaming, and reactive browser dashboards rendered via Three.js and D3.js.

< 100ms Geospatial Query Window

Sub-100 millisecond spatial radius searches across dynamic driver coordinates.

50,000+ Concurrent Telemetry Streams

High-throughput message pipeline handling continuous GPS position pings without packet queuing.

99.98% Dispatch Reliability

Resilient ride assignment state machine preventing duplicate or orphaned bookings.

60 FPS Fleet Visualizer Refresh

GPU-accelerated Three.js web view showing live driver clusters and city traffic density.

Real-Time Telemetry and Spatial Concurrency Challenges

Operating a responsive urban transportation network requires balancing high-frequency mobile GPS streams with low-latency matching algorithms and instantaneous browser visualizers.

00

High-Frequency GPS Write Volume

Thousands of moving vehicles transmitting location fixes every two seconds generated heavy write amplification and index lock contention in standard relational tables.

01

Assignment Race Conditions

Surges in local ride demand meant multiple passenger requests competed for the nearest available driver, risking duplicate dispatch assignments.

02

Large-Scale Browser Fleet Monitoring

Operations managers needed real-time visibility into entire metropolitan fleets, including vehicle statuses, headings, and speed, without freezing browser UI threads.

Distributed Geospatial Pipeline and Reactive Visualization

We architected an event-driven streaming pipeline that partitions geospatial state into in-memory spatial shards while rendering city-scale fleet dynamics with WebGL.

Spatial Sharding and Geohash Partitioning

Partitioned geographic areas using hierarchical spatial grid buckets, allowing proximity search queries to target localized memory caches rather than scanning global tables.

  • In-memory spatial index keeping neighborhood search latency under 50 milliseconds
  • Dead-reckoning motion prediction smoothing vehicle locations between cellular GPS pings
  • Adaptive ping frequencies based on vehicle velocity to reduce mobile battery drain

Deterministic Dispatch State Engine

Coordinated driver offers and passenger bookings through an event-driven state machine using optimistic concurrency tokens to guarantee single-winner assignments.

  • Idempotent message handlers preventing duplicate dispatch commands
  • Automatic fallback cascade reassigning rides when a driver declines or times out
  • Audit event log capturing every dispatch transition for operational forensics

Three.js and Canvas Fleet Dashboard

Developed an operations command center using Three.js and Canvas to render tens of thousands of active vehicle markers with instanced geometry and directional heading indicators.

  • Hardware-accelerated marker rendering maintaining 60 frames per second on desktop browsers
  • D3.js data overlays visualizing demand heatmaps and historical dispatch delays
  • Bidirectional WebSocket transport streaming filtered regional updates to connected dispatchers

Geospatial Telemetry Ingestion and Dispatch Architecture

Flow of vehicle GPS telemetry through WebSocket ingestion, in-memory spatial index sharding, matching engine, and real-time dispatcher dashboards.

100%
sequenceDiagram autonumber participant D as Driver Mobile App participant G as Telemetry Gateway (WebSockets) participant S as In-Memory Spatial Shards participant M as Dispatch Match Engine participant P as Passenger Mobile App participant O as Operations Console (Three.js) D->>G: Stream GPS Coordinate (Lat/Lng, Heading, Speed) G->>S: Update Spatial Cell Partition G->>O: Push Vehicle Position Delta (Instanced Mesh) P->>M: Request Ride (Pickup Coordinate) M->>S: Query Nearest Available Drivers (Radius Search) S-->>M: Return Ranked Driver Candidates (< 100ms) M->>D: Dispatch Ride Offer with Expiration Timer D->>M: Accept Ride Confirmation M->>P: Notify Driver En Route with Live Track M->>O: Transition Trip State to Active Dispatch

Architectural Principles Demonstrated

  • Spatial Sharding Beats Global Queries: Dividing geographic territories into memory-resident grid cells eliminates database lock contention and keeps proximity searches predictable during rush hour peaks.
  • State Machines Protect Dispatch Integrity: Treating ride assignments as strict single-winner state machines backed by idempotent events prevents driver double-booking even across distributed mobile networks.
  • GPU Acceleration Belongs in Enterprise Consoles: Offloading thousands of moving fleet markers to Three.js instanced geometry gives dispatch operators fluid control and deep situational awareness without UI stutter.
Web