Course

Lab Content 45 hours

1. Utilizing a monocular camera for object
2. Employing LIDAR for object
3. Utilizing RADAR for object
4. Object tracking via
5. Object tracking employing stereo camera
6. Measuring object distances using
7. Utilizing RADAR for object distance
8. Employing Stereo camera technology for object distance
9. Camera-based lane
10. Optical character recognition applied for detecting traffic signs

Unit 1

Introduction to ADAS: Introduction – Terminology, Design consideration, Safety assessment. Commonly used hardware, main components of software stack, Vehicle modelling and control, safety frameworks and current industry practices. State Estimation and Localization – Least squares – Vehicle localization sensors – GPS and IMU – Extended Kalman filter, unscented Kalman filter – LIDAR scan matching, iterative Closest Point Algorithm – Multiple sensor fusion for vehicle state estimation and localization.

Unit 2

Feedforward neural networks: – Review of Deep Learning, Multilayer Perceptron, Optimization, Stochastic Gradient Descent, Back propagation – Review of Convolutional Neural Networks (CNN): Architecture, Convolution/Pooling layers – Understanding and Visualizing CNN. Visual Perception – Visual Perception – Pinhole camera model, intrinsic and extrinsic camera calibration, monocular and stereo vision, projective geometry – CNNs for 2 D Object detection, Semantic segmentation.

Unit 3

Motion Planning: – Driving Missions, Scenarios, and Behaviour, Motion Planning Constraints, Objective Functions for Autonomous Driving, Hierarchical Motion Planning – Occupancy Grids, Populating Occupancy Grids from LIDAR Scan Data, Occupancy Grid Updates, High Definition Road Maps. ACC, AEBs, LDWS, LKA Creating a Road Network Graph, Dijkstra’s Shortest Path Search, A* Shortest Path Search, Motion Prediction, Map-Aware Motion Prediction, Time to Collision. UNECE, GSR, Indian regulations.

Course Objectives

1. To provide knowledge on fundamental concepts, terminologies, and design considerations in autonomous vehicle systems.
2. To make students’ understand state estimation and localization techniques including least squares, Kalman filters, and sensor fusion for accurate positioning of autonomous vehicles.
3. To familiarize students with neural networks and deep learning techniques, focusing on feedforward neural networks and convolutional neural networks (CNNs)
4. To familiarize in motion planning algorithms and techniques, including driving missions, occupancy grids, and path planning algorithms

Course Outcomes

Skills acquired

Designing and implementing state estimation algorithms, neural networks for perception tasks, motion planning algorithms, and utilizing sensor data for object detection and tracking in autonomous vehicle systems.

Text Books / References

1. Lipson, H & Kurman, M, Driverless: Intelligent Cars on the Road Ahead, MIT Press, 2020
2. Dan Simon, “Optimal State Estimation: Kalman, H∞, and Nonlinear Approaches”, John Wiley & Sons, 2012
3. Ian Goodfellow, Yoshua Bengio, Aaron Courville, “Deep Learning”, MIT Press, 2016
4. David Forsyth, Jean Ponce, “ Computer Vision: A Modern Approach”, Pearson, 2023
5. Ian Goodfellow, Yoshua Bengio, Aaron Courville, “Deep Learning”, MIT Press, 2016
6. Thrun, W. Burgard, and D. Fox, “Probabilistic robotics”, MIT Press, 2010