jackyhuynh/kalman_filter_for_localization_using_python

Kalman filters are really good at taking noisy sensor data and smoothing out the data to make more accurate predictions. For autonomous vehicles, Kalman filters can be used in object tracking. A Kalman filter does this by weighing the uncertainty in your belief about the location versus the uncertainty in the lidar or radar measurement. If your belief is very uncertain, the Kalman filter gives more weight to the sensor. If the sensor measurement has more uncertainty, your belief about the location gets more weight than the sensor measurement.

Kalman Filter

Kalman Filters - Why are they useful?

Kalman filters are really good at taking noisy sensor data and smoothing out the data to make more accurate predictions. For autonomous vehicles, Kalman filters can be used in object tracking. (Retrieved from Udaciti.com/Intro to self-Driving Car)

Kalman Filters and Sensors

Object tracking is often done with radar and lidar sensors placed around the vehicle. A radar sensor can directly measure the distance and velocity of objects moving around the vehicle. A lidar sensor only measures distance.

Put aside a Kalman filter for a minute and think about how you could use lidar data to track an object. Let's say there is a bicyclist riding around in front of you. You send out a lidar signal and receive the signal back. The lidar sensor tells you that the bicycle is 10 meters directly ahead of you but gives you no velocity information.

By the time your lidar device sends out another signal, maybe 0.05 seconds will have passed. But during those 0.05 seconds, your vehicle still needs to keep track of the bicycle. So your vehicle will predict where it thinks the bycicle will be. But your vehicle has no bicycle velocity information.

After 0.05 seconds, the lidar device sends out and receives another signal. This time, the bicycle is 9.95 meters ahead of you. Now you know that the bicycle is traveling -1 meter per second towards you. For the next 0.05 seconds, your vehicle will assume the bicycle is traveling -1 m/s towards you. Then another lidar signal goes out and comes back, and you can update the position and velocity again. (Retrieved from Udaciti.com/Intro to self-Driving Car)

Sensor Noise

Unfortunately, lidar and radar signals are noisy. In other words, they are somewhat inacurrate. A Kalman filter helps to smooth out the noise so that you get a better fix on the bicycle's true position and velocity.

A Kalman filter does this by weighing the uncertainty in your belief about the location versus the uncertainty in the lidar or radar measurement. If your belief is very uncertain, the Kalman filter gives more weight to the sensor. If the sensor measurement has more uncertainty, your belief about the location gets more weight than the sensor mearuement. (Retrieved from Udaciti.com/Intro to self-Driving Car)

Technology

• Python
• Object Oriented Design
• Jupyter Notebook
• Data Visualization
• Machine Learning
• AI
• Localization
• Prediction
• Data Structures

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a live system.

Prerequisites

What things you need to install the software and how to install them

• Jupyter Notebook: If you want just test the code, simply go to google and search for jupiter notebook or another Python online IDE. The Jupyter Notebook is an open-source web application that allows you to create and share documents that contain live code, equations, visualizations and narrative text.
• Anacoda Navigator: Install Anaconda Navigator if you want to develop data sciences using python or R. Anaconda Navigator is a desktop graphical user interface included in Anaconda that allows you to launch applications and easily manage conda packages, environments and channels without the need to use command line commands.

Installing

A step by step series of examples that tell you how to get a development enviroment running

• Install Anacoda Navigator - If you haven't downloaded and installed Anacoda Navigator yet, here's how to get started.
• Jupyter Notebook - Click here to go to the online free Jupiter Notebook.

Running the tests

Explain how to run the automated tests for this system:

• There is no download IDE need, all you need is download all the src to your machine and run it.
• Using Jupiter Notebook
• Navigate to the file Kahman_Filter_Implementation.ipynb
• hit:

Ctrl + Enter

• The notebook will execute in Markdown form and include some data visualization to show the actual performance of kalmanfilter vs. lidar vs. round truth.

Deployment

Matrices class can be deploy and ready to work with any sensor, or moving robotic prediction. Idea for localization and/or self-driving car. It turns out that using multiple sensors like radar and lidar at the same time, will give even better results. Using more than one type of sensor at once is called sensor fusion, which is used in Self-Driving Car applications. Please refer to my notebook for better understanding.

Built With

• Jupyter Notebook

Contributing

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.

Versioning

We use SemVer for versioning. For the versions available, see the tags on this repository.

Authors

• Truc Huynh - Initial work - TrucDev

Format

my README.md format was retrieved from

• Billie Thompson - Initial work - PurpleBooth See also the list of contributors who participated in this project.

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Acknowledgments

• Hat tip to anyone whose code was used
• Inspiration
• etc