1889. Random-Key Optimizers for Solving Combinatorial Optimization Problems
Invited abstract in session MC-20: Metaheuristic algorithms, stream Combinatorial Optimization.
Monday, 12:30-14:00Room: Esther Simpson 2.11
Authors (first author is the speaker)
| 1. | Antonio Chaves
|
| UNIFESP | |
| 2. | Mauricio Resende
|
| Industrial and Systems Engineering, University of Washington |
Abstract
This work presents the Random-Key Optimizer (RKO), a modular stochastic local search framework for combinatorial optimization. RKO encodes solutions as random-key vectors that are transformed into feasible solutions using problem-specific decoders. The framework integrates multiple metaheuristics: simulated annealing, iterated local search, variable neighborhood search, particle swarm optimization, genetic algorithms, large neighborhood search, and greedy randomized adaptive search procedures. These metaheuristics can operate independently or collaboratively through an elite solution pool. RKO's problem-independent architecture requires only a decoder implementation to adapt to new problems. The framework features novel continuous-space local search heuristics using randomized variable neighborhood descent that combines an adapted Nelder-Mead with other optimization techniques. Implemented in C++ with multi-threading capabilities, RKO enables solution sharing among concurrent metaheuristics. We demonstrate its effectiveness on four NP-hard problems: the tree of hubs location problem, traveling thief problem, portfolio problem, and node-capacitated graph partitioning problem. Results confirm RKO's robustness and adaptability as an effective tool for combinatorial optimization. The RKO solver is publicly available on GitHub.
Keywords
- Metaheuristics
- Combinatorial Optimization
- Software
Status: accepted
Back to the list of papers