The Study Of Mobile Robot Path Planning Based On Improved Genetic Algorithm

Mobile robot is a kind of robot which can work in the complex environment withself-planning, self-organizing and adaptive ability. The research on mobile robot beganin1960s,and the robotics became more and more advanced and comprehensive afternearly half a century of development, the robots have been used increasingly in a widerange of area, they can not only be used in the field of welding, cargo handling to reduceproduction costs, but also can accomplish those tasks which are too dangerous for humans,suchas the disposal of nuclear waste, scientific exploration and study deep-sea resources. Therefore, therobot technology has been widespread concern worldwide, and became an important symbol ofnational science and technology capacity levels.Intelligent optimization algorithm in path planning has its unique advantagescompared with traditional planning algorithm, can adapt to the dynamic environmentand make decisions like humans. Genetic algorithms due to its own characteristics, hasa few limit for the target problem, deals with the problem flexibly, and good at solvingcomplex problems and nonlinear problems, with good implicit parallelism and globalsearch capability. So GA gets widespread attention in the mobile robot path planning,and has made a series of achievements, but it is easy to fall into local optimal andpremature convergence.This paper combines the grid method to build the environment model with improvedgenetic algorithm to path planning, and studies the mobile path planning separately inthe static environment and dynamic environments. At first, using the grid method tobuild the model, and encoding by the serial number, then searches the free grid so thatthe initial population is viable path without barrier grid or intermittent path. Whenconducting genetic manipulation, this paper uses variable length chromosomeencoding based the grid number, designs fitness function static and dynamic fitnessfunction, static fitness function mainly considers the path length, while dynamicfitness function considers the path length, whether the robot is in the safe range and dynamic collision avoidance these three factors. In addition to three basic operators--the selection, crossover and mutation operator, but also to optimize the design of thethree operators: Insert operator for the intermittent path into a continuous path; deleteoperator is used to remove redundant paths point to reduce the length of the path;smooth operator to optimize the path to make it smoother. This paper also designsanother three optimization operator: the insert operator turn path into a continuouspath; delete operator is used to delete redundant path points to reduce the length of thepath; smooth operator is used to optimize the path to make it smoother.This paper set up two different environment models and made simulations. Madecomparison with simple genetic algorithm in static simulation environment; thedynamic environment is divided into two kinds, the first on is with added stationaryobstacles, the second on is with moving new obstacle. Conducted simulationexperiments and statisticed experimental data. The feasibility and effectiveness of thealgorithm have been proved, and shows the improved genetic algorithm can findfeasible path by a higher rate, and the path is much shorter and smoother, collision-free,proved the superiority of the algorithm.