Improvements Of ACO And PSO Algorithms And Their Applications To Logistics Distribution

Logistics distribution is a modern and effective logistics style, so it has been widely applied in modern business circulation.And because of computer technology's development, the modern logistic has been improved a lot.Under normal circumstances, distribution route can be described as follows: there are N client points and the demand quantity and location of each point are known; there are at most K vehicles which are used to send the goods from delivery center to each client point; after finishing the delivery, all the vehicles need to return to the center. Every vehicle's load is fixed, and finds a shortest vehicle running path which can make the least transportation costs.According to different research focuses , logistic distribution can be divided into: single logistics center problem and multi logistics center problem; infinite time problem and finite time problem; vehicle-open problem and vehicle-closed problem; point-to-point transportation, multi-point transportation, Single-loop transportation-TSP, Multi-loop transportation.Whether the distribution path is reasonable or not has huge effect on distribution speed, cost and effectiveness. When we are searching for the optimized method,our goal is to achieve the optimized goal required by the customers.For the moment, there are a lot of solutions for the research and development to solve the VRP which can be divided into two major categories: accurate algorithm and functional algorithm. As the VRP is the NP-hard problem, the more the number of the customers are, there will more solutions for the path way, even to the immeasurable scale. And in the real life, most of the VRP are based on large data volume, which direct the research methodology of the article to focus on such swarm intelligence like ACO and PSO in order to construct the heuristic algorithm of high-qulity.ACO was first successfully applied to solve logistics distribution path optimization problem, such as the traveling salesman path. The research showed that: ants can release pheromone when they are searching for food. Pheromone can be used to communicate and cooperate with other ants in order to find a shorter path. The more ants go through one area, the more concentration the pheromone will be and finally nearly all the ants will choose this shortest path. This is the positive feedback phenomena. It uses distributed parallel compute mechanism and is easy to be combined with other methods. It also has robustness. Its disadvantage is that searching time is long and apt to fall into local optimal solutions. PSO originated from the evaluation of the birds'moving and social during the feeding.Every bird is called a particle. Every particle has a adaptable value determined by a optimized function. For each particle, there is an attribution-speed which determines the direction and distance of particle and then all the particle follow the current optimal particle to search in the solution space. PSO algorithm is a evolutionary computation technology based on swarm intelligence. This technology introduces the concept"swarm"which is a optimized tool based on swarm. Compared with Genetic Algorithm,the advantage of PSO is simple, easy to achieve and has not many parameters to adjust.The paper makes full use of match point of ACO and PSO in swarm algorithm and makes the ant which has searched all the city nodes and had path information as the new particle, converts the path information based on expression style and then makes it as the particle initial solution. According to the particle swarm ideology, for each particle, in accordance with the fly direction and distance, the solution is to be searched in solution space following the current optimized particle and optimal individual. For characteristics of large-scale data, Cross-division is used (cross-group mechanism) to improve the particle swarm and the low accuracy and divergent solutions can be avoided. Based on this, we use optimized solution which has been iteratively adjusted by PSO to update the pheromone of ACO and make the particle which has path information as the new ant to initialize the path. Then the ant and the particle's study ability can be improved. The better solution can be obtained when there are fewer ants and to some extent, the local optimal solution can be avoided.Optimization algorithm's main strategies are:1. the continuity problem of the particle swarm has been converted to vehicle path's discrete problem effectively and when the PSO is combined with the ACO, there is a good interface.2. parameter optimized mechanism.3. when the initial solution is formed, optimization algorithm makes the path traversed by the ant as the initial solution of particle swarm and then the particle in particle swarm has more guidance. For ACO,the particle in the PSO will search better solution based on the one calculated by the ACO, and effectively avoid the local optimal solution problem which is apt to appear in ACO.4. this optimized algorithm uses Cross-division mechanism. The initial solution has more guidance for it makes use of ant's better solution. So, we use the city scale 10 as the dimension of the particle and use the path searched by ant to initialize the particle. Because the city is small, it is easier to find a better solution.5. In the improved optimized algorithm, the pheromone of the path is updated by ant path which has been dealt with by PSO, so it has more guidance.Through the test results for 10 times based on the ten groups of experimental data, we find that when the number of the cities is 50, the number of the ants in the combined algorithm is less than that in the ACO by 50%, but the lengths of the best route and the route draw are improved by 8% and 8.5%. when the number of the cities is 100,150,1000, the number of the ants in the combined algorithm is less than that in the ACO by nearly 50%, but the lengths of the best route and the route draw are improved by 8% and 15% Experiments show that the hybrid optimized algorithm can make better path solution searched around the certain area in ACO. Based on four group of experiment data, through many tests repeatedly, it is proved that this method can find out better solution with updating single individual's optimal solution and overall optimal solution. Meanwhile, through updating the pheromone by the path adjusted by ACO, the ants do have more guidance and efficiency. Through compare, we can find that, in the finite iterative process, under the circumstance of fewer ants, the hybrid optimized algorithm can effectively lower the defect of the original algorithm-be apt to fall into local optimal solution, improve the quality of the optimal solution and convergence rate a lot.