Does AlphaEvolve support solving constrained combinatorial optimization problems?

[wumian0929]

Hello, I would like to ask if AlphaEvolve can support the solution of constrained combinatorial optimization problems? Such as vehicle routing problems with time window constraints and network planning?

[codelion]

Yes, OpenEvolve can absolutely support constrained combinatorial optimization problems like VRPTW and network planning!

Key Requirements for Success

1. Automated Evaluation Function: You need a function that can automatically score solutions while penalizing constraint violations (capacity limits, time windows, etc.) rather than rejecting infeasible solutions outright.

2. Initial Working Solution: Provide at least one baseline heuristic that produces valid (even if suboptimal) solutions - like nearest neighbor, savings algorithm, or sweep algorithm.

3. Constraint Handling Strategy: Use penalty methods where constraint violations add cost to the objective function, allowing the evolution process to gradually learn to satisfy constraints while optimizing the primary objective.

4. Problem Representation: Structure your problem so that solutions can be represented as code that OpenEvolve can modify and evolve.

What OpenEvolve Will Discover

The evolutionary process will discover novel hybrid approaches that combine:

• Different route construction methods
• Local search improvement operators
• Constraint handling strategies
• Problem decomposition techniques