What factors can I model with cuOpt? Can I model distance traveled or carbon emissions?

NVIDIA cuOpt is designed to solve vehicle routing problems (VRPs) using cost and travel time matrices. These names reflect how the solver interprets the data internally—but they don’t limit what you can actually model. In practice, cuOpt is flexible about what those numbers represent.

How cuOpt interprets inputs

At a high level, cuOpt supports:

• Travel time and cost matrices defined at the vehicle type level
• Maximum limits on time and cost that can be applied at the individual vehicle level

These are the official labels you’ll see in the cuOpt product. However, cuOpt does not enforce semantic meaning on these values beyond how they are used in the optimization.

What this means for modeling

You are free to use these matrices to represent any additive quantity relevant to your routing problem, such as:

1. Distance traveled
2. Carbon (CO₂) emissions
3. Fuel consumption
4. Driver effort or fatigue scores
5. Risk, noise, or any other measurable impact

For example:

• You might minimize operational cost while treating CO₂ emissions as the “travel time” input.
• You could use distance as cost and time as a hard constraint.
• You could encode a composite score (e.g., distance × emissions factor) as the objective.

cuOpt will optimize exactly what you provide—regardless of the label.

Tracking metrics without optimizing them

Sometimes you want to track and report a metric without letting it influence the optimization. NVIDIA cuOpt supports this pattern as well for a single factor.

If you assign a weight of 0 to the cost component cuOpt will still accumulate and report the value, but the metric will not affect the objective function. This is useful when you want to answer questions like:

• What are the total emissions of the optimized routes?
• How much distance did we travel, even though we optimized for cost?

Key takeaway

Although cuOpt uses the terms cost and travel time, these are best thought of as generic numeric channels. As long as your values are additive across routes, cuOpt can handle them—whether they represent time, money, distance, emissions, or something entirely different.

If you’re unsure how to structure a specific modeling approach, feel free to reach out—we’re happy to help you map your real-world objectives into cuOpt’s inputs.