Programming a Robotic Car

In the last few weeks I have been quite busy following the course "Programming a Robotic Car" on Udacitity. The course is quite demanding because it force you to implement all the algorithms presented. I think this is by far the most effective way to really learn and their on-line platform based on Python works pretty well.

These are some of the problems that can be solved with the techniques explained:

• Localizing a robot in a map
• Monte Carlo Localization
• Particle Filters Localization
• Tracking objects
• Kalman Filters
• Motion planning
• A Star
• Dynamic Programming
• Robot Motion
• Smoothing
• PID Control
• Simultaneous localization and mapping
• SLAM
• Online SLAM

You can access all the code in my public repository under the following path:

• Udacity >> CS 373 - Programming a robotic car

Here few snippets of the most relevant code. 

Kalman Filter:

Particle Filter:

A Start:

Smoothing:

PID Control:

Twiddle (used for tuning the PID control parameters):

Online SLAM:

Robot Localization using Montecarlo Localization

The problem of localising a robot in a known environment is quite interesting. I would like to share with you my solution to the homework assignment for the unit 1 of the "CS 373 Programming a Robotic Car".

The environment is a discrete map (colours matrix) where each area can be green or red. The robot can sense the colour (sense function) of the area where is currently on and can move in any directions (move function). To make things more interesting and close to reality the measurements and the motions are subjected to errors and this is represented by the variables sensor_right and p_move. Given a list of measurements and a list of motions the problem is to create a matrix of probability that describe where the robot is likely to be after all the specified moves.

More information about the definition of the problem can be found on Udacity.

I used the python module unittest for this purpose. This is one of the unit tests that I defined for the problem:



This is the main code that does the actual calculation:

The current algorithm used to solve this problem is called Montecarlo Localization.

You can find the full code in my personal repository.