Application Of An Improved Particle Swarm Optimizer Algorithm In Path Planning

With the development of artificial intelligence, the application of robots (such as:Industrial, agricultural, medical, aerospace.etc.) is more and more widely in modernlife recently. Path planning is the most basic part for the robots to implement alloperations, whose essence is to find a path, which is satisfied some certain criteriaand has no collision with all obstacles from the start to the end, in the environmentfull of obstacles according to some criteria (such as: the shortest path, less timeconsuming, security and so on). In this task, this problem is discussed from thefollowing aspects:(1) Environment of optimization: Generally, the environment for path planningusually contains static environment and dynamic environment. Static environmentdescribes that obstacles don’t change with time, while obstacles vary with thechange of time in dynamic environment. This task is based on the basis to studythe general process of path planning under static environment.(2) Process of problems: There is always not only a optimized criterion for pathplanning problems and security and the shortest path are regarded as main criteriawin this topic. These two criteria can be taken as a whole objective, which is atotal penalty function to evaluate the problem. In the previously work, a staticpenalty function is used to describe the problem, while dynamic constraintmethods are employed to overcome the drawback of static penalty function inlater period. On the other hand, multi-objective optimization is applied to dealwith this kind of problem for more than one evaluation. Because security and theshortest path are contradictory and can’t be satisfied at the same time, somemethods should be proposed to meet the demand of different users, so thesecriteria are regarded as multi-objective problem in later period work.Multi-objective optimization could optimize a few objectives simultaneously andCoordinate these goals, and the algorithm will generate a group of Pareto optimalsolutions every time for the designer, so user can select the suitable solutionsaccording to their need for criteria. (3) Selection of curve: To ensure the generated path smoothness is critical conditionin the process of during the robot navigation for path planning. Most ofresearchers are concentrated on finding the shortest path, least cost of time,security and so on, which always make the generated path discontinuous orunsmooth, so find a suitable curve to describe path is of significance. There aremany curves to generate path, such as: Bezier curve, Ferguson curve,3curve etc.Bezier curve, Ferguson curve and3curve are compared through the application ofdifferent curves in path planning, then a conclusion about a more suitable curvefor this problem can be get, which could be used to describe path in latterapplication.(4) Methods for path planning: There are many methods for path planning, containingtraditional method grid method, artificial potential field method etc. In recentlyyears, heuristic algorithms and artificial intelligence algorithm are graduallyapplied to path planning. Particle Swarm Optimizer (PSO) has achieved verygood results in all kinds of optimization problems on multi-dimensionalcontinuous space for its small number of individuals, simple computing, strongglobal optimization, fast convergence speed, good robustness etc. features. Inorder to overcome the drawback of standard Particle Swarm Optimizer, animproved Particle Swarm Optimizer–Dynamic Multi-Swarm Particle SwarmOptimizer (DMS-PSO) is proposed to employ in mobile robot path planning inthis task.