Research On Path Planning And Mapping For Mobile Robot

Robotics is one of the most active areas in high technology research.The research of mobile robots is an important branch of robotics.It involves many disciplines and has been widely applied in various fields.Autonomous navigation is the basic function of mobile robots.Path planning and mapping are the keys to make a robot navigate and explore environment autonomously.Path planning is to determine a collision-free path between the start and target position in an environment with obstacles by a performance criterion.As the base of path planning and motion control,mapping is to model environment by data acquired from sensor.SLAM(simultaneous localization and mapping)combines robot localization and mapping as an estimation problem.By SLAM,robots can get more reliable environment map.It is necessary for a mobile robot in autonomous navigation to locate itself,build environment map and plan a path.It is of great theoretic and practical significance to study path planning and mapping for mobile robot.In this dissertation,path planning and SLAM for mobile robots are studied.The main contents of this dissertation are described as follows.1.The impact of outlier disturbance is ignored for the traditional SLAM algorithms based on Kalman filter.The outlier disturbance may make a robot be unable to locate itself.In this dissertation,an anti-outlier disturbance KF-SLAM algorithm is proposed to solve this problem.This algorithm introduces outlier disturbance detection and covariance inflation to EKF-SLAM(extended Kalman filter SLAM)and UKF-SLAM(unscented Kalman filter SLAM).The anti-outlier disturbance method improves estimation accuracy and robustness.2.Incorrect priori knowledge about the control and observation noise matrices would seriously degrade the accuracy of the FastSLAM algorithm.An improved FastSLAM algorithm based on DFC&ASD-PSO(dynamic fractional calculus and alpha-stable distribution particle swarm optimization)is proposed.An extra step is introduced to the FastSLAM 2.0 framework to adjust the noise matrices.This step is implemented by the inconsistency between time-adjacent observations and DFC&ASD-PSO.By the more accurate estimation on the priori knowledge,it improves the accuracy of SLAM.3.To overcome the problems of particle depletion and the linear approximations of thenonlinear functions,an improved FastSLAM algorithm based on SR-UKF(square root unscented Kalman filter)and revised genetic resampling is proposed.In this algorithm,the robot pose is updated by SR-UKF.As SR-UKF propagates the Sigma points through the true nonlinearity,it decreases the linearization errors.By directly transferring the square root of the state covariance matrix,SR-UKF has better numerical stability.To amend particle degeneracy and keep particle diversity,this algorithm combines double roulette wheels as the selection operator,fast Metropolis-Hastings(MH)as the mutation operator and traditional crossover to a novel genetic algorithm to resample particles.4.In global path planning algorithms based on visibility graph,the visibility graph construction is very time-consuming and visibility graphs include many useless edges.To improve efficiency of global path planning,this dissertation proposes a novel global path planning algorithm,SVGA(simultaneous visibility graph construction and path optimization by A*).This algorithm does not construct a visibility graph before the path optimization.However it constructs a visibility graph and searches an optimal path at the same time.Only edges which are related to the optimal path are added to the visibility graph by the heuristic search,and most edges are ignored.The performance analysis validates that SVGA algorithm can find the optimal path and reduce the computational cost.5.Visibility graph is only used in a known environment.To overcome this limitation,this dissertation proposes a novel path planning algorithm based on modified visibility graph which involves cubic B-spline curves and particle swarm optimization.This algorithm consists of three steps,dynamic polygon generation,path searching and path smoothing.Trapping in local minima and discontinuities often exist in local path planning.To help robots escape from traps,the environment is memorized by combining polygons in the dynamic polygon generation process.To solve discontinuities,the path is smoothed by cubic B-spline curves and particle swarm optimization(PSO)algorithm.This smooth path is more adapted to the kinetics constraint of mobile robots.