joyyy821 / pf-slam

Particle filter SLAM and texture mapping based on differential-drive robot.

ECE276A - Project 2: Particle Filter SLAM

Author: Yang Jiao

Date: March 7, 2023

Local Python environment:

• Python version: 3.10.9
• Libraries: numpy, matplotlib, opencv-python

Mapping results

5 Particles	100 Particles

Note that the animation speed does not reflect the program executing speed. The program runs significantly slower with 100 particles than with 5 particles.

See the full SLAM results and texture mapping results in project report.

Project Hierachy

The submitted .zip contains all of files in the code folder. main.py is the entrance for executing the project. It invokes class methods from models.py file which are the implementation for particle filter SLAM and texture mapping. The file sensors.py is modified based on the given load_data.py file and the RGBD image conversion code snippet from piazza. Both sensors.py and models.py invoke provided functions from pr2_utils.py file.

-----code
|    |-- readme.md
|    |-- main.py
|    |-- sensors.py (Load data from data folder)
|    |-- models.py (SLAM and texture mapping implementations)
|    |-- pr2_utils.py (provided code)
|
|----data
     |-- dataRGBD (folder with all RGBD images)
        |-- Disparity20
        |-- Disparity21
        |-- RGB20
        |-- RGB21
     |-- Encoders20.npz
     |-- Encoders21.npz
     |-- Hokuyo20.npz
     |-- Hokuyo21.npz
     |-- Imu20.npz
     |-- Imu21.npz
     |-- Generated SLAM and texture mapping results will also appear here

Executing the Code

Below are the instruction to execute this project.

1. Extract the .zip file and copy all .py files to the code folder in a same structure of project workspace as shown in Project Hierachy section;
2. Open the root folder (containing both code and data) in a Python IDE or cd to the root folder in terminal;
3. Run main.py file in the IDE or use terminal command such as python ./code/main.py.
   • One can (remove) comments on specific codes in the main.py file to execute different tasks;
   • One can test on different dataset by passing different arguments to initialize the Robot() class instance;
   • To modify the noise variance, need to go to the models.py file and modify the initParticles(self, N) method in the Robot() class;
4. The meaning of different mode is stated at the beginning of the __init__() method for the Robot() class, which is also pasted here:

   # mode:
   # 0 - dead reckoning
   # 1 - prediction only
   # 2 - particle filter
   # 3 - texture mapping