This post will show you how I progress in my task of subject “Robótica de Servicios” to program an amazon warehouse robot:

Amazon Warehouse

This task presents us with the situation of a robot in an Amazon warehouse which can lift and put down shelves. The robot has to use OMPL to get the path to the known points where the shelves are located, reach these points following the path and lift the shelves and bring them to the central point where it was located.

Planning the Implementation

The method that I have decided to implement is a sequence system with diferents states.

First, the robot get the path, then it follows it and arrive to the shelve. Then it turn and lift the shelve and follow the path back to get to the origin position and put down the shelve in there after stoping

Method Implementation

In this practice I will perform substask to convert to gaxebo coordenates to pixels and the absolute2relative function used before, also I used the OMPL class.

Used Libraries

The code libraries that I have used are time:

• I have used numpy.
• I have used math.
• I have use opencv to get the path and also os.path and sys libraries needed by OMPL.

Code Functions

For this task I needed to use two functions.

The first function, called “absolute2relative”.

• Absolute2relative function:

    def absolute2relative (x_abs, y_abs, robotx, roboty, robott):
        dx = x_abs - robotx
        dy = y_abs - roboty
  
        x_rel = dx * math.cos (-robott) - dy * math.sin (-robott)
        y_rel = dx * math.sin (-robott) + dy * math.cos (-robott)
          
        return [x_rel, y_rel]
  

The second function, called “convert_to_pixels”, is a function that is responsible to return in the pixels the corresponding coordinates in the real world.

• Convert to pixels function

    def convert_to_pixels(x_real_world, y_real_world):
        proporcion_x = (-x_real_world + high_real_world / 2) / high_real_world
        proporcion_y = (-y_real_world + width_real_world / 2) / width_real_world
      
        x_pixels = proporcion_x * high_pixels
        y_pixels = proporcion_y * width_pixels
  
        return int(x_pixels), int(y_pixels)
  

Critical Analysis

Finally, I have to mention that to get the path I used the 2D point planning using a PPM image as a map in which an image is used as a map and defines the robot as a point, also I used opencv instead of ppm so I have to convert the code to opencv, the original code with ppm is in this link.

Also I have to mention that to follow the path I used a lower velocities because if the velocities are high enough, this will pass the point and to recover it it must make a complete turn causing it to collide with the surrounding objects and be lost.

Videos

Simulation Video

This video shows the simulation of the robot going to the first shelve and move it to the origin point.