The Traveling Thief Problem with Time Windows: Benchmarks and Heuristics

Researchers Helen Yuliana Angmalisang and Frank Neumann have published a new paper introducing and benchmarking the 'Traveling Thief Problem with Time Windows' (TTPTW). This complex optimization problem combines three real-world constraints: a thief must travel between cities (like the Traveling Salesperson Problem), steal items with varying weights and values (the 'Thief' component), and do so within specific time windows when targets are accessible. The paper, submitted to arXiv, presents this as a more realistic benchmark for AI and algorithms than studying these components in isolation.

To evaluate solutions, the team created new TTPTW benchmark instances based on existing TTP literature. They tested adaptations of established approaches for the classic Traveling Thief Problem and the Traveling Salesperson Problem with Time Windows against this new challenge. Their experimental results, detailed across 13 pages, show that a newly designed heuristic algorithm developed by the authors consistently outperforms the other adapted methods across a wide range of these benchmark scenarios. This work advances the field of combinatorial optimization and provides a valuable new testbed for AI systems designed for real-world scheduling and routing tasks.