On the Use of Iterative Problem Solving for the Traveling Salesperson Problem with Changing Time Window Constraints

A research team led by Hy Nguyen has published a paper demonstrating that iterative problem-solving significantly improves optimization for real-world logistics challenges. Their work focuses on the Traveling Salesperson Problem with Time Windows (TSPTW), a complex routing problem where deliveries must occur within specific time windows. The key insight is that many real-world routing problems are similar—think daily delivery routes with slightly changing customer time windows—and knowledge from previous optimization runs can be reused rather than starting from scratch each time.

The researchers created a multi-task benchmark where each base routing instance was expanded into five related tasks under two realistic scenarios: partial time-window expansion (like customers extending their delivery windows) and swap-additive time reassignment (where time constraints shift between customers). They compared a standard from-scratch approach against their iterative protocol, which initializes each new task's solution using the best tour from the previous task. Using three popular optimization algorithms—LNS (Large Neighborhood Search), VNS (Variable Neighborhood Search), and LKH-3—their results showed the iterative method was consistently superior in progressive-relaxation settings and generally competitive under swap-additive changes, with improvements increasing on more difficult problem instances.

This research bridges a gap in optimization literature by systematically comparing sequential transfer learning against independent task solving for TSPTW. The practical implication is substantial: logistics companies, ride-sharing services, and supply chain operators can implement this relatively simple protocol to get better solutions faster when faced with sequences of related routing problems. As the authors note, the approach is particularly valuable when travel-time matrices are fixed but time-window constraints change across tasks—exactly the situation faced by companies with regular delivery routes facing daily schedule adjustments.