summer-sessions-2019

A series of lectures and hands-on tutorials organized between the 6th of May 2019 and 20th of June 2019 to familiarize new lab entrants with the fundaments of different areas of robotics research that the lab conducts.

An important aspect of this session will be getting familiar with various robots and sensors present in the lab. This knowledge/skill will be critical as some of the assignments will involve using these systems to collect data and/or execute the algorithms.

Venue and Timings:

• Venue : Nilgiri 119 (Saranga Hall)
• Timings : 15:00 Hrs to 17:00 Hrs

Tentative topics:

Lab Robots & other Hardwares: ROS (Robot Operating System), Cameras(Monocular, Stereo, RGBD), Lidars, IMUs, Flight controllers, Husky Robot, P3DX Robot, Bebeop Drone, GPS, etc.

Fundamentals of Linear Algebra, Calculus and Optimization: Vectors, Matrices, Vector and Matrix operations, Important matrices, Matrix decompositions, Gradient, Hessian, Linear and non-linear least squares, Unconstrained optimization methods, Lagrange multipliers.

Rigid Body Transformations: Rotation matrices, Homogeneous Transformation matrices, Rigid Body transformation, Composition of transformation by current-axis and fixed-axis conventions.

Geometric methods in Computer Vision: Projective geometry, Camera modelling, Camera Calibration, Two-view geometry, Triangulation, Resection, SfM, Visual odometry.

Deep learning for Computer Vision & Tensor Flow

Motion Planning: Robot modelling, Motion Planning overview, Sampling based planning, Variational methods for planning.

Reinforcement Learning: Markov Processes, Planning using Dynamic Programming, Value Iteration and Policy Iteration, Model Free RL(TD learning, SARSA, Importance Sampling), table Q-learning, Value Function Approximation(Incremental and Batch), Deep Q Networks(DQN), Policy Gradients(REINFORCE, Actor-Critic Methods). (Additionally we can also cover Deep RL(DQN, Double DQN, Asynchronous DQN, Bayesian RL, TRPO, PPO, Inverse RL etc.) if we have time).

Schedule

Date	Topic	Presenter	Material	Additional Reading
6-5-19	Intro to lab and sessions, Maths review	Mithun	Class Slide	Matrix_Differentiation_Reference
7-5-19	Maths review contd.	Mithun	Class Slide	Reference Materials
9-5-19	Rigid Body Transformations and Image Formation	Gourav	Class Slide Image Formation	at end of slide
11-5-19	Rigid Body Transformations contd.	Gourav	Class Slide	at end of slide
13-5-19	Camera Calibration	Junaid	slide	Camera Intrinsics and Extrinsics
15-5-19	Multiple-view Geometry 1 - Overview, Intro to visual odometry, Feature detection and matching, Motion estimation	Karnik	Class slides SIFT, F-matrix slides	Original SIFT paper MVG Ch. 9 & 10
17-5-19	Multiple-view Geometry 2 - Epipolar geometry, RANSAC	Karnik	Class slides	MVG Ch. 9 & 11 MVG Sec. 4.6 Moving object detection paper
20-5-19	Stereo	Mahtab	TBD	TBD
22-5-19	Multiple-view Geometry 3 - Triangulation, Resection, Bundle Adjustment	Karnik	Class slides	MVG Ch. 12 E-PnP Bundle adjustment
23-5-19	DL 1	Sarthak	Basic review of ML and forward propagation	TBD
24-5-19	DL 2	Shashank	Back Propagation, CNN & Optimization Methods	CS231n lecture slides
27-5-19	DL 3	Sarthak	PyTorch Introduction and Coding	TBD
28-5-19	DL 4	Shashank	CNN Architectures, RNNs/LSTMS, object detection	CS231n lecture slides
29-5-19	DL 4	Shashank	Coding LSTMs and time series prediction	TBD
1-6-19	Motion Planning	Mithun	Motion Planning overivew and Graph search methods	Slides, Slides
3-6-19	Motion Planning	Mithun	Sampling based methods and Local Planning	Slides
8-6-19	Trajectory Generation/ ROS	Mithun, Gourav	ROS Tutorial	ETH ROS Course
9-6-19	ROS	Gourav	ROS Tutorial	ETH ROS Course
19-6-19	RL Basics	Kaustubh	slides	DeepMind Lectures

Assignments

Assignment No.	Release Date	Topic	files	Deadline
1	10-5-19	Linear Algebra, Optimization, Transformations	Problem statement	14-5-19
2	15-5-19	DLT Calibration	Problem statement	19-5-19
3	25-5-19	Two-view reconstruction	Problem statement	29-5-19
4	27-5-19	Deep Learning_Assign1	Problem statement	3-6-19

References

Linear Algebra / Vector Calculus / Optimization:

• Refer Issue #1 and Issue #2

Multiview Geometry:

• Photogrammetry II - A course by Prof. Cyrill Stachniss
• Multiple View Geometry in Computer Vision - A book by Richard Hartley and Andrew Zisserman. Colloquially referred to as the bible.
• An Invitation to 3D Vision by Yi Ma, Stefano Soatto, Jana Kosecka, and Shankar S. Sastry - Considered more beginner-friendly than the bible.
• Photogrammetric Computer Vision by Wolfgang Förstner and Bernhard P. Wrobel - The book Cyril Stachniss follows in his lecture series.

Motion Planning:

• Refer Issue #13

Reinforcement Learning:

• Book: Reinforcement Learning by Sutton and Barto pdf
• Reinforcement Learning lectures by David Silver link

Contacts:

• Maths(Linear Algebra, Optimization, Probability & Statistics)

Mithun Nallana [email protected]

• Rigid Body Transformations, Image Formation, Camera Geometry

Gourav Kumar [email protected]

• Projective Geometry, Multiview Geometry

Karnik Ram [email protected]
Junaid Ahmad [email protected]

• Deep Learning

Sarthak Sharma [email protected]
S Shashank [email protected]

• Reinforcement Learning

kaustubh mani [email protected]

• Motion Planning and Trajectory Optimization

Mithun Nallana [email protected]

• Hands-on sessions with Robots and sensors

Sriram N N [email protected]
Gourav Kumar [email protected]

Note: Announcements will be made through the repo's issues page, so please subscribe to the repo (by clicking the 'watch' button) to receive notifications.