Problem with navigation and move_base stacks

edit

I just want to check the efficiency of gmapping module and designed a little experiment.

I have created a 3D environment(a maze) using blender and imported it into Gazebo by creating an sdf with the mesh produced in blender as visual and collision properties. Then I created a Gazebo world with the maze(mesh), light source and ground_plane.

Parameters in ground plane:

static - true
link - 'link'
  collision - 'collision'
    geometry -> plane -> normal(0 0 1), size(500 500)
  visual - 'visual'
    geometry -> plane -> normal(0 0 1), size(500 500)

Then I used a launch file which launches the above world in Gazebo and run move_base, gmapping, explore_lite, amcl, map_server(node) and RViz.

Everything seems to be good to me. But there is a problem with bot`s position during the course of exploration. For example, bot goes from (0,0) to (5,0) in Gazebo and in the map shown in RViz as well. But then suddenly in the next moment bot shifts few steps back(like from (5,0) to (3,0)) in the RViz map but stays at the same position in Gazebo as it should be. And as the program keeps running this sudden jumps become more frequent and map gets destroyed.

Initially, I thought this is problem of inefficient localisation. So, then I ran same launch file, but this time I removed the amcl package. Annoyingly, the problem remains the same.

Then I tried running the same process with gui(both Gazebo and RViz) to see if this is because of low computation power. The problem still presists.

I`m using turtlebot, ROS Kinetic on Ubuntu 16 with Gazebo7. This whole experiment is being done on Quadcore HP Laptop with i3 processor with Max clock speed of 1.7GHz and min clock speed of 0.8MHz. Clock Speed during this whole process is almost 1.7GHz even without rendering GUI.

can anyone please help me with this problem.

Is this because of inappropriate definitions in the world file or low processing power.