Multimodal Fleet Optimization: Routing Trucks, Vans, and Cargo Bikes as a Single Optimization Problem
Most logistics operations still route each vehicle type in isolation. Trucks get one set of routes, vans another, and cargo bikes (if they exist at all) are treated as a side project. That approach leaves money on the table. When a 3.5-ton truck delivers a single parcel to a pedestrian zone, or a cargo bike sits idle while a van crawls through congested city centers, you are paying for inefficiency baked into separate planning workflows.
Multimodal fleet optimization treats trucks, vans, and cargo bikes as a single pool of capacity, then assigns shipments to the right vehicle based on payload, distance, access constraints, and cost. Done correctly, it cuts fuel spend, shrinks delivery windows, and improves driver utilization without adding headcount.
Trucks, vans, and cargo bikes are planned together, and each shipment goes to the right vehicle based on payload, distance, access constraints, and cost.
The engine calculates total cost, not just distance or time.
Access constraints go in as hard limits, not soft penalties, or the algorithm proposes illegal routes.
It hinges on how much your constraints differ from the standard model.
Why Traditional Routing Breaks Down with Mixed Fleets
Legacy transportation management systems were built when fleets were homogeneous. A carrier ran trucks or vans, not both. Route optimization algorithms assumed uniform capacity, speed, and cost per mile. Adding cargo bikes or electric micro-vehicles to that model creates edge cases the solver cannot handle.
You end up with three separate route plans. Dispatchers manually shuffle orders between vehicle types when reality does not match the plan. That manual layer introduces delay, error, and lost visibility.
The problem compounds in dense urban markets where access restrictions, low-emission zones, and parking scarcity make vehicle choice as important as the route itself. A truck banned from a city center at peak hours is useless, no matter how efficient the algorithm.
Separate plans per vehicle type
Three separate route plans- Trucks get one set of routes, vans another
- Algorithms assume uniform capacity, speed, and cost per mile
- Dispatchers manually shuffle orders between vehicle types
- The manual layer adds delay, error, and lost visibility
One multimodal optimization
A single pool of capacity- Trucks, vans, and cargo bikes planned together
- Vehicle type and route solved simultaneously
- Each shipment matched on payload, distance, access, and cost
- Total cost, not just distance or time
What Makes Multimodal Optimization Different
Multimodal fleet optimization solves for vehicle type and route simultaneously. Instead of asking “what is the best route for this truck,” the system asks “what is the lowest-cost way to deliver these 200 parcels using any combination of trucks, vans, and cargo bikes.”
That requires modeling several new constraints:
- Vehicle-specific access rules: cargo bikes can enter pedestrian zones, trucks cannot.
- Payload and volume limits
- Cost per kilometer by vehicle type
- Service time variability
- Depot and micro-hub locations: cargo bikes reload from neighborhood micro-hubs, not the main depot.
The solver must balance these trade-offs in real time. A 50 kg parcel going 2 km into a congested zone might cost less on a cargo bike, even though a van is faster on paper. The optimization engine calculates total cost, not just distance or time.
Figure 1. One solver chooses the vehicle and the route
Restates this article: orders, the vehicle pool and the constraints go into one solve, which returns routes for every vehicle type as one plan and re-solves when a vehicle breaks down or a high-priority order arrives. Structure only; no values are drawn.
How the Math Changes
Classic vehicle routing problems (VRP) minimize distance or time for a homogeneous fleet. Multimodal VRP adds a discrete choice variable for vehicle type at each stop, turning the problem into a mixed-integer program. Solvers like Google OR-Tools, VROOM, or Gurobi can handle this, but the formulation matters. You need to encode access constraints as hard limits, not soft penalties, or the algorithm will propose illegal routes.
Start with a two-tier approach:
- Cluster orders by geography and access constraintsFor example, city center vs. suburbs.
- Assign vehicle types to clustersBased on aggregate payload and cost.
- Run detailed routingWithin each cluster, using the assigned vehicle type.
That heuristic runs fast enough for daily planning. For real-time adjustments, you need agentic AI for autonomous route recalculation in logistics that can re-solve on the fly when a vehicle breaks down or a new high-priority order arrives.
Implementation: Build or Buy?
You have three paths:
- Extend your existing TMS: If you run a modern transportation management system, check whether it supports multi-vehicle routing. SAP TM and Oracle OTM do, but you will need professional services to configure constraints and cost models.
- Adopt a specialized multimodal platform: Tools like Bringg, Urbantz, and OptimoRoute focus on last-mile optimization with native support for mixed fleets. Faster to deploy, but you may need API integration to connect with your ERP and order management systems.
- Build a custom solver: If your fleet is large or your constraints are unusual (e.g., temperature-controlled cargo bikes, cross-docking between vehicle types), a custom solution built on OR-Tools or Gurobi gives you full control.
The build option makes sense when off-the-shelf tools cannot model your specific cost structure or when you need to embed optimization into a broader AI-powered supply chain analytics platform.
What to Watch: Regulation and Infrastructure
Two external factors will accelerate multimodal adoption:
- Low-emission zones
- Micro-hub proliferation: Cities are converting parking spaces and vacant storefronts into neighborhood consolidation points.
If your market has either trend, waiting another year to optimize multimodal routing means ceding cost advantage to competitors who moved earlier.
Key Takeaways for Engineering and Product Leaders
Multimodal fleet optimization is not a science project. It is a cost reduction lever that works today, provided you have clean order data, vehicle telemetry, and a solver that can handle mixed-integer constraints. Start with a pilot in your densest urban market, measure cost per delivery before and after, and scale from there.
If you are evaluating whether to extend your TMS, buy a new platform, or build custom, the decision hinges on how much your constraints differ from the standard model. Off-the-shelf works when your fleet is trucks, vans, and standard cargo bikes. Custom makes sense when you need to model cross-docking, multi-temperature zones, or real-time re-optimization at scale.
Sthenos helps logistics and e-commerce companies build and integrate multimodal routing engines that cut costs and improve delivery performance. If you are ready to move beyond isolated route planning, we can show you what a unified optimization model looks like for your fleet.
Related services: AI and machine learning development; SAP services.
Related reading
Multimodal routing in depth
- Multimodal Fleet Optimization: Assigning Cargo Bikes, Vans, and Trucks in Real-Time
- Multi-Modal Fleet Optimization: One Engine, All Vehicle Types
- Multi-Constraint Route Optimization in Real-Time Logistics
The terms behind it
- What Is Route Optimization?
- What Is Last Mile Delivery?
- What Is a Transportation Management System (TMS)?
