COIMBATORE INSTITUTE OF TECHNOLOGY

(Autonomous Institution Affiliated to Anna University)

II M.Sc. AIML – IV Semester

19MAM42-Machine learning

Case Study Presentation

Work done by: Keershaa .V.S

Roll number: 71762234024

Date: 18.3.2024

CASE STUDY TITLE: AUTONOMOUS SELF DRIVING CARS

INTRODUCTION

• Autonomous self-driving cars, also known as driverless or robotic cars, are vehicles
capable of navigating roads and making driving decisions without human intervention.
• These cars are equipped with advanced sensors, cameras, radar, LiDAR, and artificial
intelligence (AI) algorithms to perceive their surroundings, interpret traffic signs, and
react to obstacles in real-time.
• The development of autonomous driving technology represents a significant milestone in
transportation innovation, offering potential benefits in terms of safety, efficiency, and
accessibility.
Importance and Relevance of the Problem

• Traditional human-driven vehicles are associated with various risks and challenges,
including human error, accidents, traffic congestion, and environmental impact.
• Autonomous self-driving cars have the potential to address these challenges by:
• Improving Safety: By eliminating human error, autonomous cars can reduce the
number of accidents and fatalities on the roads.
• Enhancing Efficiency: Autonomous vehicles can optimize traffic flow, reduce
congestion, and minimize travel times through advanced routing and
coordination
• Increasing Accessibility: Autonomous driving technology can provide mobility
solutions for individuals who are unable to drive due to age, disability, or other
limitations.
• Reducing Environmental Impact: By promoting more efficient driving patterns
and facilitating the transition to electric and shared vehicles, autonomous cars
can contribute to reducing greenhouse gas emissions and combating climate
change.

Dataset Description and Characteristics

• The dataset used for this case study consists of real-world sensor data collected from
autonomous vehicles during test drives.
• It includes a diverse range of attributes, such as:
• Sensor Readings: Data from cameras, LiDAR, radar, GPS, and other sensors
mounted on the vehicles.
• Vehicle Dynamics: Information about speed, acceleration, steering angle, and
other vehicle parameters.
• Environment Characteristics: Weather conditions, road types, traffic density,
and other environmental factors.
• Size: The dataset contains thousands of samples collected over multiple test sessions,
resulting in a large volume of data.
• Dimensionality: Each sample in the dataset is high-dimensional, with multiple attributes
and features captured by the sensors.
• Variability: The dataset captures various driving scenarios and conditions, including
urban, highway, and rural environments, as well as different weather and lighting
conditions.
Data Preprocessing Steps:
1. Cleaning:
1. Identify and handle missing values, outliers, and errors in the dataset.
2. Remove or impute missing values using techniques such as mean, median, or
interpolation.
3. Detect and remove outliers that may skew the analysis or affect model
performance.
2. Normalization:
1. Scale and standardize numerical features to ensure they have a similar range
and distribution.
2. Normalize sensor readings and vehicle parameters to facilitate model
convergence and improve performance.
3. Feature Engineering:
1. Extract relevant features from raw sensor data to capture meaningful
information.
2. Create new features based on domain knowledge or insights from exploratory
data analysis.
4. Handling Categorical Variables:
1. Convert categorical variables, such as weather conditions or road types, into
numerical representations using one-hot encoding or label encoding.
5. Dimensionality Reduction:
1. Apply dimensionality reduction techniques, such as principal component
analysis (PCA) or feature selection, to reduce the complexity of the dataset and
improve computational efficiency.
6. Data Splitting:
1. Split the dataset into training, validation, and test sets to evaluate model
performance and prevent overfitting.

SUMMARY STATISTICS
 The fuel efficiency of self-driving cars is projected to increase 10-20%
 By 2040, 70% of new cars are expected to have some form of automation
 Autonomous taxis would lessen the need for parking by 68%
 62% of people are willing to pay extra for self-driving car technology
 39 million autonomous vehicles are expected to be on the road globally by 2030
 Excessive speed is the cause of 25% of fatal car accidents, which could be reduced by
self-driving cars
DATA PREPROCESSING

DATA VISUALIZATION
HISTOGRAM:
Path Evaluation & Visualization

How automotive Artificial Intelligence algorithms are used for

self-driving cars

Tasks are classified into 4 sub-tasks:

1. Perception
2. Localization
3. Prediction
4. Decision Making
 I. PERCEPTION:
To achieve such a high level of perception, a self-driving car must have three sensors:

Camera
LiDAR
RADAR

CAMERA:

LIDAR

RADAR
II.LOCALIZATION:

III.PREDICTION:
The car has a 360-degree view of its environment that enables it to perceive and capture all the
information and process it. Once fed into the deep learning algorithm, it can come up with all
the possible moves that other road users might make. It’s like a game where the player has a
finite number of moves and tries to find the best move to defeat the opponent.
The sensors in self-driving cars enable them to perform tasks like image classification, object
detection, segmentation, and localization. With various forms of data representation, the car can
make predictions of the object around it.
A deep learning algorithm can model such information (images and cloud data points from
LiDARs and RADARs) during training. The same model, but during inference, can help the car
to prepare for all the possible moves which involve braking, halting, slowing down, changing
lanes, and so on.
The role of deep learning is to interpret complex vision tasks, localize itself in the environment,
enhance perception, and actuate kinematic maneuvers in self-driving cars. This ensures road
safety and easy commute as well.
IV.DECISION MAKING:
The Machine Learning Algorithms Used in Self-Driving Cars:
• SIFT (scale-invariant feature transform) for feature extraction

 AdaBoost for data classification

  TextonBoost for object recognition

 Histogram of oriented gradients (HOG)

 YOLO (You Only Look Once)

Model Training :

1. Data Preparation:
1. Split the dataset into training, validation, and test sets.
2. Preprocess the data, including normalization, feature scaling, and encoding
categorical variables.
2. Model Selection:
1. Choose appropriate machine learning algorithms for the task, such as deep
learning models (e.g., CNNs, RNNs), reinforcement learning, or traditional
supervised learning models (e.g., decision trees, SVMs).
3. Hyperparameter Tuning:
1. Select optimal hyperparameters for the chosen model, such as learning rate,
batch size, and regularization parameters.
 2. Use techniques like grid search, random search, or Bayesian optimization to find
the best hyperparameters.
4. Training the Model:
1. Train the model on the training data using the selected algorithm and
hyperparameters.
2. Use techniques like mini-batch gradient descent, Adam optimization, or other
optimization algorithms to update the model parameters iteratively.
5. Monitoring Training Progress:
1. Monitor training progress by tracking performance metrics on the validation
set.
2. Use techniques like early stopping to prevent overfitting and save the best-
performing model checkpoint.

EVALUATION METRICS:
1. Performance Metrics:
1. Evaluate the model's performance using appropriate metrics for the task, such
as accuracy, precision, recall, F1 score, mean squared error (MSE), or area
under the ROC curve (AUC).
2. Cross-Validation:
1. Perform cross-validation to assess the model's generalization performance
across different subsets of the data.
2. Use techniques like k-fold cross-validation or stratified cross-validation to ensure
robust evaluation.
3. Test Set Evaluation:
1. Evaluate the final trained model on the test set to obtain an unbiased estimate of
its performance.
2. Compare the model's performance on the test set with its performance on the
training and validation sets to detect overfitting.
4. Visualization and Interpretation:
1. Visualize model predictions and errors to gain insights into its behavior.
2. Analyze misclassified samples or prediction errors to identify areas for
improvement.
5. Comparative Analysis:
1. Compare the performance of different models or algorithms to select the best-
performing approach.
2. Consider factors such as computational complexity, scalability, and
interpretability in addition to performance metrics.
CONCLUSION:

Autonomous self-driving cars promise safer, more efficient transportation. Challenges include
technical complexity, regulatory issues, and societal acceptance. Continued research and
collaboration are essential for realizing their full potential and navigating toward a future of
enhanced mobility

REFERENCES:
 https://mindy-support.com/news-post/how-machine-learning-in-automotive-makes-self-
driving-cars-a-reality/
 https://neptune.ai/blog/self-driving-cars-with-convolutional-neural-networks-cnn