Esnaf-007: A Life-Like Negotiation Agent
This study considers a routing problem with a heterogeneous fleet. The fleet size is not limited, and demands are 2D vectors. Thus, this study cares additional contraints for demands unlike the classical vehicle routing problems. The problem is modeled mathematically, and a feasible solution is obtained with Cplex solver. In this study, heuristics are built to get close feasible solutions to Cplex in more reasonable time. Various approaches are developed to investigate their behaviour and parameter tuning routines. Two different constructive heuristics are created, and their effect to improvement heuristics as an initial solutions are analyzed. There are also three different improvement heuristics, to improve an initial solution. Well-known first improvement and best improvement strategies are analyzed in the study. The stochastic approaches while accepting non-improving solutions to prevent stuck are also investigated with simulated annealing algorithm. Finally population based heuristics are also evaluated. Two different graph representation is built to solve the problem. A genetic algorithm is also introduced for future studies but not completed. These study contributes to see heuristic approaches to deal with unusual constraints in an optimization problem. Experiments also show that why heuristic approaches should be used to deal with such complex problems that considerably close, even better, objective values are obtained in significantly less time compared to an exact solver. Moreover, results also show that tuning parameters has great importance and this study provides some pre-knowledge about parameters' effects on performance.
KEYWORDS: heuristic algorithms, NP-hard, VRP, heterogeneous fleet
Reach the full project details here