RouteForge — Geospatial Logistics Optimization Engine
The Problem
System Bottlenecks & Technical Friction
Last-mile logistics networks operating across dozens of metro zones struggle with exponential route permutation complexity. Dispatchers often rely on manual spreadsheets or off-the-shelf routing tools that cannot accommodate complex multi-dimensional constraints: union break mandates, vehicle height/weight restrictions, dynamic time windows, and real-time traffic congestion. Suboptimal dispatching causes excessive driver overtime, missed delivery windows, and costly rescue runs.
Our Engineering Approach
Architecture Design & Implementation
RouteForge is a constraint programming solver built on Google OR-Tools' vehicle routing problem (VRP) framework with custom constraint modeling. The solver models 18 hard constraints (time windows, vehicle capacity, shift limits, mandatory rest breaks, road restrictions) and 6 weighted business objectives (distance minimization, workload balancing, highway preference). PostGIS stores geocoded spatial layers, while Kafka streams live status events to trigger real-time Celery re-optimization tasks.
Technical Architecture
System breakdown & stack.
Google OR-Tools CP-SAT constraint programming solver with custom VRP formulation
18 hard constraints: time windows, capacity, shift limits, breaks, vehicle-street compatibility, cross-docks
6 soft constraints weighted by business priority: minimize distance, balance load, prefer highways, reduce idle time
PostGIS spatial database with geocoded addresses, traffic matrices, metro zone polygons, street restriction layers
Mapbox Directions API for traffic-adjusted drive time matrices, refreshed every 15 minutes via Celery beat
Kafka event streaming: driver mobile app status updates → Celery re-optimization triggers
Redis route cache: pre-solved common delivery clusters for <2-second fallback retrieval
React dispatch dashboard with Mapbox GL JS: live driver tracking, route replay, ETA deviation alerts, rescue dispatch
Engagement Timeline
Engineering delivery schedule.
Constraint modeling workshop with dispatch team, business rule documentation, solver feasibility research
OR-Tools VRP formulation, custom constraint implementation, synthetic data testing
PostGIS database schema, geocoding pipeline, Mapbox Directions API integration, drive time matrix build
Kafka event streaming, Celery async re-optimization triggers, Redis route cache layer
React dispatch dashboard, Mapbox GL JS driver tracking, route visualization, ETA accuracy metrics
Parallel-run validation (OR-Tools vs. manual planning), solve time optimization, production cutover
More Case Studies
Related AI projects
Want results like these for your business?
Book a 30-minute scoping call with a senior engineer. We'll scope your AI project, define the architecture, and give you a fixed-price proposal within 5 business days.