Práctica 4: Global Navigation using Teletaxi

-

  

Welcome!

Picture 1. Taxi in New York

Goal:

In this exercise we are going to implement the logic of a Gradient Path Planning (GPP) algorithm, which consists of:

- Selecting a destination, the GPP algorithm is responsible for finding the shortest path to it, avoiding, in the case of this practice, everything that is not road.

- Once the path has been selected, the logic necessary to follow this path and reach the objective must be implemented in the robot.

Explanation:

First, we have to built a gradient map assigning weights to every cell between the beginning and the target we want to reach.

I have choose 2 different values for the weights: 1 for cardinal directions (north, south, east and west) and 1.25 for the intermediate directions (north-east, south-east, north-west and south-west).

Next, we have to binarize the map (binarize_map() method) from where we obtain only 2 values, 0 if it´s an obstacle and 255 if it´s road. All the data obtained is copied into a numpy array of zeros.

Then, we use the previous numpy array to fill the gradient (using the fill_gradient() method), where I have used a Queue to pop or push the values in the gradient depending if they are or not obstacles in the way. 

We always put the lower values in the array. If we have reached our objective, we add some extra iterations to give the car some more maneuvering time so that it can hit the target correctly.

With the obstacles list obtained in the fill_gradient() method we have taken into account all directions and corners in the map.

Next, we have to get a path, and for this task I have used the path_extraction() method, and with their values assigned using the assign_weights() method we iterate from the initial point to the target we want to reach in the map.

Finally, to make the car move, I have used a similar method from the 3rd practice of the subject (obstacle avoidance), where I have used only attractive forces, no repulsion forces are needed here because the obstacles are obtained in the fill_gradient method().

MULTIMEDIA CONTENT (SIMULATION IN UNIBOTICS):

Image: View of the map and the gradient

Picture 3(1). Environment visualization


Picture 3(2). Environment visualization

Video: Execution of the program

Picture 4. Execution of the program

Comentarios

Publicar un comentario

Entradas populares de este blog

Práctica 1: Vacuum Cleaner

-
Imagen
Welcome! Figure 1. Roomba vacuum cleaner Goal: In this exercise we are going to implement the logic of a navigation algorithm for an autonomous vacuum, where the main objective will be to cover the largest area of a house using the programmed algorithm. Explanation: In the first place I have created the function parse_laser_data(), which is attached to the statement of the practice, and whose main functionality is to read the data that the laser collects and with which we will play when deciding the next behavior of our vacuum cleaner. The operation of my vacuum cleaner is based on a state machine, which consists of 4 states: - State 0 (spiral): This state is only performed once in execution, the vacuum cleaner performs a spiral to cover the entire possible area around its initial position, until it detects an obstacle (this is done by reading the values from the front of the laser), and once it detects the obstacle it goes to the next state. - State 1 (reverse): The vacuum cleaner goe...
Leer más

Práctica 2: Follow Line

-
Imagen
  Welcome! Figure 1. Formula 1 car Goal: In this exercise we are going to perform a PID reactive control (represented in a Formula 1 car) capable of following the red line painted on the middle on the road of the racing circuit. Explanation: In this exercise we were asked for a program that would allow a F1 car to travel the circuit as fast as possible using image filters and a controller from the PID family. In my case I have opted for a PD controller, whose formula is as follows:                     Picture 2. Controller used in this exercise When I was doing it using this method I have realized something that has seemed quite curious to me, and that is that after numerous attempts giving values to the constants of each controller (Kp and Kd) I have realized that the oscillation of the car is null when both constants have the same value. I have considered the linear speed constant, varying it between a range of 2 and 2.5...
Leer más

Práctica 3: Obstacle Avoidance

-
Imagen
  Welcome! Picture 1. Overtaking in an F1 race Goal: In this exercise we are going to implement the logic of the VFF (Virtual Force Field) navigation algorithm, where each object generates a repulsive force towards the robot and it´s destiny an attractive force. Explanation: In this exercise we were asked to design a version of the VFF (Virtual-Force-Field) algorithm, in which an F1 car must travel the circuit taking into account that this time it will have obstacles on the way and you will have to dodge them. First, I've declared three constants, one to check if you've hit the target, and the other two corresponding to the alpha and beta values ​​that will be applied to calculate the resultant force. Then I have implemented the function absolute2relative(), which, as its name indicates, converts the absolute coordinates (point (0,0) at the beginning of the circuit) to relative ones (positive x-axis means forward and positive y-axis means left). The functions corresponding to t...
Leer más