Memetic algorithms for evaluating post-disaster scenarios

Natural disasters often cause severe damage to transport networks, which poses critical challenges to humanitarian logistics, especially in the rapid and accurate assessment of the needs of the population in affected areas. Gathering information is vital to prioritizing relief efforts and avoiding wasted resources. To mitigate this problem, this work focuses on the Combined Drone Orienteering Problem (CDOP), an NP-Hard problem that coordinates the joint use of drones and trucks to maximize information collection under severe time and battery life constraints. In CDOP, the truck acts as a mobile base, allowing the drone to visit shelter sites, such as communities, schools, or hospitals, that have become inaccessible or dangerous for ground vehicles due to blockages and damage to the road network.
A Memetic Algorithm organized in a ternary tree structure with 13 agents is proposed. Each agent maintains a pair of solutions: Current, focused on exploring the search space, and Pocket, a memory engine that preserves the best individual solution found, ensuring that the leader of each subpopulation is always superior to their children. The method integrates a genetic algorithm with domain-specific knowledge through the optimization procedure, which combines a 'Education' step of the routes with insertion-based Local Search algorithms. Warm Start uses cheapest insertion heuristics for the truck and incremental profit maximization for the drone, allowing you to start the evolutionary process from higher quality solutions