What we build
Production routing — wired into the dispatch console.
Each capability is a production component — not a proof-of-concept — wired to your TMS, telematics, and driver apps, documented for ISO 9001 and labour-law expectations, and monitored continuously.
Multi-constraint VRP solver
OR-Tools and metaheuristic search honour vehicle capacity, time windows, driver shifts, and cooling rules at the same time. Cold-chain, mixed-fleet, and pickup-and-deliver patterns supported out of the box.
Live re-optimisation
Traffic, weather, and exception events trigger partial re-solves on the fly. Only the affected legs reshuffle — drivers see a 2-tap accept on the next stop, not a confusing whole-route refresh.
Day-of + next-day planning
Tonight's batch solve covers tomorrow's full plan. The day-of engine takes over once vehicles roll, absorbing late orders, customer reschedules, and live road conditions without a planner in the loop.
Cost + CO₂ telemetry per route
Cost-per-delivery and CO₂-per-stop attributed against route, hub, vehicle class, and customer segment. Sustainability and finance teams pull the same ledger your dispatchers run on.
Driver-app integration
Turn-by-turn handoff to Onfleet, Bringg, or your own Android / iOS app. Roster sync, geofenced check-ins, proof-of-delivery capture, and exception flags loop straight back into the planner.
Hub + yard sequencing
Loading-dock waves, gate scheduling, and cross-dock windows planned alongside vehicle routes. Yard congestion and bay assignments feed the same solver, so dispatch and dock dance to one beat.
Who we route for
One engine, every fleet motion.
Same VRP solver, ETA predictor, and dispatch console — tuned per network and route family. Shared feature store, per-hub adapters, per-shipper SLA bands that match the contract you actually signed.
3PL last-mile carriers
B2B and B2C parcel networks running mixed-fleet last-mile across metros and tier-2 cities. Per-hub dispatch consoles, network-wide rollups, and SLA reporting per shipper.
Quick-commerce + dark stores
10-30 minute SLA fulfilment from dark stores and ghost kitchens. Dynamic batching, courier-pool balancing, and live re-routing as orders compound through the lunch and dinner peaks.
Cold-chain F&B distribution
Chilled and frozen distribution with temperature-band, reload-window, and dwell-time constraints. Reefer telemetry feeds straight into the solver so spoilage risk is treated like a hard constraint.
Field-service fleets
Utilities, telco, and appliance-repair fleets with appointment windows, parts-inventory constraints, and skill-matched technicians. First-time-fix rate optimised alongside drive-time.
Returns + reverse logistics
E-commerce returns and dropoff aggregation tied back into outbound rounds. Returns are routed alongside deliveries when capacity allows, cutting empty miles and reverse-leg cost.
Cross-border + multi-leg freight
SG-MY-TH-VN regional moves with customs windows, ferry schedules, and driver-handover points. Route plans honour border-crossing windows and downstream linehaul connections.
Model families we deploy
No single model explains every route. So we compose.
Solving, ETA prediction, exception triage, and disruption forecasting each have their own model family — composed into one dispatch pipeline with version control at every step.
OR-Tools and metaheuristic large-neighbourhood-search engine that handles VRPTW, VRPPD, and mixed-fleet variants. Reinforcement-learning-tuned objective weights — SLA, cost, drive-time, CO₂ — with tight reproducibility on every solve.
ML model that predicts arrival windows from telematics, traffic, and historical dwell features. Per-driver, per-zone, per-time-of-day calibration produces P90 bands customers and dispatchers can rely on — retrained nightly on closed routes.
ML classifier scores live exceptions — failed delivery, mis-load, customer no-show, traffic incident — by severity and recommends the next action. Routes the alert to driver, dispatcher, or customer-comms automation accordingly.
ML forecaster predicts volume shocks and route degradation from weather feeds, event calendars, and historical disruption patterns. Lets the planner pre-stage capacity before the storm — not react to it afterwards.
Data sources wired into every route
Every feed that moves the truck — integrated.
Pulled in parallel, normalised into a single dispatch-ready schema, reconciled live against the TMS, and audit-logged alongside the solve version that produced the plan.
Explainability, not just routes
Every route carries its reasoning. For the dispatcher. For the driver.
Every solve, ETA, and re-route is accompanied by objective weights, input provenance, and a plain-language reason a dispatcher can paste into a customer message — generated at solve time, indexed for audit.
- Objective weights cited on every solve
- Input provenance traced to the TMS feed
- Re-opt reason logged for every disruption
- Override audit sealed on every manual change
Frameworks we align to
Why Axccelerate for route optimisation
Not a routing vendor.
A logistics operating system.
A vendor gives you a route map. Our stack gives you solving, ETA prediction, live re-opt, dock scheduling, exception triage, and CO₂ telemetry — the infrastructure a real dispatch desk actually runs on.
Pricing
Priced to the network footprint, not the stop count.
Routing deployments are custom — we scope against your hubs, vehicle classes, and SLA bands before quoting.
Glossary
The vocabulary behind every dispatch decision.
A quick reference for the acronyms that show up in route optimisation — the terms your dispatchers, planners, and ops director will all use.
- VRP
- Vehicle routing problem
The combinatorial-optimisation problem of assigning a set of stops to a set of vehicles so that capacity, time, and other constraints are honoured at minimum cost. The mathematical core of route optimisation.
- VRPTW
- VRP with time windows
The variant of VRP where each stop has a delivery window. Most last-mile and field-service problems are VRPTW underneath.
- VRPPD
- VRP with pickup and delivery
VRP variant covering jobs that pair a pickup with a delivery on the same vehicle — courier, ride-pooling, and intra-city freight typically run on VRPPD.
- LNS
- Large-neighbourhood search
A metaheuristic that destroys and re-builds parts of an existing solution to escape local minima. The workhorse for high-quality VRP solutions at production scale.
- OR-Tools
- Google's optimisation library
An open-source toolkit of constraint-programming and routing solvers maintained by Google. The reference engine for VRP work that needs to ship in weeks, not quarters.
- Time window
- Stop delivery window
The earliest and latest acceptable arrival time at a stop. Time windows are the difference between a 'cheap' route and one your customer is willing to pay for.
- ETA model
- Predicted arrival window
A model that produces an expected arrival time and confidence band for a stop, fed by historical telemetry, traffic, and dwell features. Quoted to the customer and used internally for SLA budgeting.
- Geofence
- Virtual perimeter
A virtual boundary around a hub, customer, or zone. Geofences trigger automatic check-ins, dwell logs, and arrival events without requiring driver-side button presses.
- Dwell time
- Time stationary on-stop
The duration a vehicle spends at a stop. A core input into the ETA predictor and a key signal for hub and dock-scheduling decisions.
- Cold-chain
- Temperature-controlled logistics
Logistics where products must remain inside a defined temperature band end-to-end. Reefer cabins, reload windows, and door-event audits are all cold-chain disciplines.
- Cross-dock
- Hub transfer without storage
A logistics pattern where freight moves from inbound truck to outbound truck through a hub without entering long-term storage. Tight dock and yard scheduling is non-negotiable.
- Last-mile
- Final-leg delivery
The final leg of a parcel's journey from a depot to the recipient. Most route-optimisation effort and most operating cost lives here.
- Linehaul
- Long-distance leg
The long-distance trunk leg between hubs. Linehaul plans interlock with last-mile so trucks roll out of the hub on the right wave.
- CO₂-per-stop
- Emissions per delivery
Estimated emissions attributable to a single stop on a route. Calculated from drive segments, vehicle class, and load — feeds finance, sustainability, and customer reporting.
Your dispatch desk, engineered.
30-minute scoping with a senior engineer and a logistics-operations specialist. You'll leave with a hub map, telematics plan, and realistic timeline — not a sales pitch.