The optimization problem in logistics
Modern logistics involves thousands of interdependent decisions made under tight time constraints: which driver covers which stops, in what order, using which vehicle, departing from which depot, subject to time windows, vehicle capacities, driver hours-of-service limits, and real-time traffic. These are all combinatorial optimization problems — NP-hard in theory, but tractable at scale with the right hardware.
Last-mile delivery routing
Optimize daily delivery runs for parcel carriers and e-commerce fulfillment. Handle time windows, package weight limits, and driver shift constraints across hundreds of stops.
Fleet dispatch & rebalancing
Assign vehicles to demand zones in real time. Re-optimize as orders arrive, cancellations occur, and traffic conditions change — with sub-minute solve times.
Warehouse pick-path optimization
Sequence warehouse picker routes to minimize travel distance across aisles. Typical deployments see 15–25% reduction in order pick time.
Load planning & cargo packing
Pack shipments into containers and trucks to minimize wasted space and minimize the number of trips. 2D and 3D bin packing with weight and axle-load constraints.
Hub network design
Determine the optimal number and location of distribution hubs to minimize total transportation cost across a supply network. Facility location as a QUBO.
Why GPU solvers for logistics?
Traditional VRP solvers (OR-Tools, LKH, CPLEX) are single-threaded or limited in parallelism. They work well for static, pre-planned routes — but struggle with real-time re-optimization as orders arrive throughout the day. NEROX GPU Annealing runs thousands of independent optimization threads simultaneously, producing high-quality routes in seconds rather than minutes, enabling truly dynamic dispatch systems.
Getting started
See the Vehicle Routing and TSP problem pages for code examples and API documentation. Contact us for a logistics-specific onboarding session with your own data.