New TTP-D hybrid AI solver optimizes drone-truck collections with 2 benchmark wins

Logistics operators face a classic dilemma: vehicles slow down as they accumulate payload, and adding a drone for outlying collections creates complex synchronization challenges. The new Travelling Thief Problem with Drone (TTP-D) formalizes this scenario, maximizing profit net of time-based rental costs while jointly optimizing item selection, vehicle routing, and flight launch/rendezvous points. Authors Kabir Murjani and Abhay Sobhanan tackle the problem with three complementary approaches: a mixed-integer linear program for small instances, metaheuristics for large ones, and an attention-based Deep Reinforcement Learning (DRL) policy for scalable speed.

The standout contribution is a learner-initialised hybrid solver where DRL constructs an initial solution, then a short annealing run refines it. On two benchmark sets, this hybrid recovers most of the metaheuristic baseline's quality at a fraction of its computational budget—though the largest instances still require full-budget metaheuristics. Sensitivity analysis reveals the rental ratio (cost per time unit) is the primary profit driver, while fleet parameters like drone speed or capacity affect margins only slightly. For professionals in route optimization, TTP-D bridges the gap between academic models and real-world drone-truck fleets, offering a practical path to faster, cheaper collection operations.