Data-driven-motion-planning-using-various-machine-learning-algorithms

Introduction

The objective of the project is to perform path planning for a four-wheel vehicle with nonholonomic constraints, using Machine Learning, I shall use different ML Models like LInear Regression and DNN to predict the outputs properly. After checking the perfomance of both the models, the one with a better result will be finalized to train with more data and will be tested and optimized. The model is being trained and tested on real-time data that has been captured using Laser scan.

Personnel

Bharadwaj Chukkala
UID: 118341705
Bharadwaj Chukkala is currently a Master's student in Robotics at the University of Maryland, College Park, MD (Batch of 2023). His interests include Machine Learning, Perception and Path Planning.
Contact me at:

Contents

├───bchukkal_final_project_report.pdf
├───bchukkal_final_project_presentation.pptx
├───LICENSE
├───README.mdta
├───codes
|   ├───data_preprocessing.py
|   ├───feature_engineering.py
|   ├───finalized_model.py
|   ├───linear_regression_model.py
│   └───neural_network_model.py
├───Data
│   ├───Test Dataset
│   └───Train Data set
├───image
│   └───README
└───results

Requirements/Dependencies

To run this project, you will need the following:

• Python 3.6+
• TensorFlow python3 -m pip install tensorflow
• Scikit Learn pip install -U scikit-learn
• Matplotlib python -m pip install -U matplotlib
• NumPy pip install -U numpy
• Pandas pip install -U pandas
• Training Data
• Test Data

Setup

1. Clone this repository to your local machine

   git clone https://github.com/bharadwaj-chukkala/Data-driven-motion-planning-using-various-machine-learning-algorithms.git
   cd Data-driven-motion-planning-using-various-machine-learning-algorithms
   

2. Install the required libraries using pip install
3. Download the dataset and place it in the data directory, there are already existing datasets present in the repository which you can choose to change.
4. The Data directory contains two folders Train Data Set and Test Dataset.
5. Both the dataset folders contain two csv files each.
   1. The Train Data Set contains two csv files July22_69 and Training_set_merged. Here the former is raw unprocessed data and the latter is clean and processed data on which we train our model.
   2. The Test Dataset contains two csv files July22_68 and Test_set_merged. Here the former is raw unprocessed data and the latter is clean and processed data on which our model makes predictions.

Instructions

Pipeline to be followed

• Data Preprocessing
• Feature Engineering
• Training different models:
  • Linear Regression model
  • Neural Network model
• Finalizing the model
• Prediction Testing
• Hyperparameter Tuning
• Regularization
• Generalization

Data Preprocessing

1. Open the codes directory and open the data_preprocessing.py and give the path to the datasets as per the need.

2. run the data_preprocessing.py file

3. The preprocessed data wil be generated in the root directory.

   cd codes
   code .                         #Opens the folder in VSCode
   python3 daata_preprocessing.py
   

Feature Engineering

1. Open the codes directory and open the feature_engineering.py and give the path to the pre processed datasets as needed.

2. run the feature_engineering.py file

3. The extracted data will then be genrated in the root directory

   cd codes
   code .                          #Opens the folder in VSCode
   python3 feature_engineering.py
   

Training the Models

Open the codes directory

1. Linear regression

   • Open the linear_regression_model.py specify the path to the engineered dataset
   • Run the training script using python3 linear_regression_model.py

2. Neural Network

   • Open the neural_network_model.py specify the path to the engineered dataset
   • Run the training script using python3 neural_network_model.py

Testing with the Final Model

1. Once the model is trained, you can use it to make predictions on new data. Open the codes directory
2. Open the finalized_model.py and specify the path to the engineered test dataset
3. Run the test script using python3 finalized_model.py

Results

The performance of the trained model will be evaluated using metrics such as accuracy, precision, and recall.

Figures

• Cross-correlation heat map

  

• Standard deviation chart of laser

  

• $E_{in}$ vs $Epochs$

  • Linear Regression

    

  • Deep Neural Network

    

• Learning Curves [Train and Test]

  

• Hyperparameter Tuning

  

• Regularization

  

Conclusion

We have successfully followed the machine learning pipeline to develop a model for estimating the ’command actions’ for the car-like robot model. The trained model is computationally reasonable by decreasing the number of features considered and a proper generalization has been shown as indicated by the results.

---

Thank you for using this project.

Copyrights 2022: @bharadwaj0chukkala