Perception in Robotics course, at Skoltech, MS in Data Science, during T3, 2022. About us: we are the Mobile robotics Lab. at Skoltech

This repository includes all material used during the course: Class notes and problem sets.

• Instructor: Gonzalo Ferrer
• Teaching Assistant: Konstantin Pakulev
• Teaching Assistant: Hekmat Taherinejad
• OFFICIAL Telegram channel with TAs

For the logistics, this course will be taught in hybrid mode, that is, the student will watch the corresponding class in advance (video) and then other related activities will happen, such as Q&A, seminars and small exercises to strengthen the learning experience.

For each lecture, all material will be included in the folder L*, and you will find the class notes, the handwritten notes as a result of the class and a short exercise we will do in class. We recommend you to print the class notes and we strongly recommend to write your own notes on the printed document while following the video class.

We will use the classes recorded from the last year (see our previous class20, class21) and we have redesigned the related material to each class for general discussion, such as exercises and further explanation of the methods discussed.

Link to the YouTube channel with video lectures

• L01: The Expectation Operator
• L02: Gaussians
• L03: Gaussian II
• L04: Bayes Filter and Kalman Filter
• L05: Motion and Sensor Models
• L06: EKF and Localization
• L07: Particle Filter and Monte-Carlo Localization
• L08: EKF SLAM with known correspondences
• L09: Data Association
• L10: Smoothing and Mapping (SAM), GraphSLAM
• L11: Pose SLAM
• L12: 3D Poses and Rigid Body Transformations
• L13a (Optional): Differentiation over SE3
• L13b (Optional): Optimization on SE3
• L14: Point Cloud Alignment
• L15: Mapping
• L16: Introduction to Visual SLAM

Problem Sets

Deadline dates for submitting problem sets, in the folder PS*:

• 9-Feb-2022, PS1: Gaussians and Visualization
• 20-Feb-2022, PS2: Localization
• 22-Feb-2022, PS2: Localization NEW DATE
• 10-March-2022, PS3: SLAM NEW DATE

Final Course Project

Final project. Teams of 3 students solving an open project. The final project could be either of the following, where in each case the topic should be closely related to the course:

• An algorithmic or theoretical contribution that extends the current state-of-the-art.
• An implementation of a state-of-the-art algorithm. Ideally, the project covers interesting new ground and might be the basis for a future conference paper submission or product.

You are encouraged to come up with your own project ideas, yet make sure to pass them by Prof. Ferrer before you submit your abstract

Logistics:

• Ideally 3 students per project (the scope of multi-body projects must be commensurate).
• Proposal: 1 page description of project + goals for milestone. This document describes the initial proposal and viability of the project.
• Progress: 3 page milestone due. You are not graded on the milestone. Think of it as a sanity check for yourself that you indeed have started to make progress on the project and an opportunity to get feedback on your progress thus far, as well as on any revisions you might have made to your project goals.
• Presentations: The presentation needs to be 12 minutes long; There will be a maximum of 3 minutes for questions after the presentation.If your presentation lasts more than 12 minutes, it will be stopped. So please make sure the presentation does not go over.
• Paper: This should be a IEEE conference style paper, i.e., focus on the problem setting, why it matters and what is interesting/novel about it, your approach, your results, analysis of results, limitations, future directions.Cite and briefly survey prior work as appropriate but do not re-write prior work when not directly relevant to understand your approach.
• Evaluation: Each team will evaluate their colleagues’ presentations.Templates will be provided the presentation day. All these points will be summed for a final evaluation (30% of the total grade).

Project proposals:

1. Comparison of outlier filtering methods such as AdaLam, Lowe ratio test and ORB-SLAM's Adaptive-FH in terms of their influence on pose estimation quality. (Reserved)
2. Comparison of DFE with conventional 8-point algorithm inside a RANSAC on pose estimation. Assesing generalization ability of DFE and testing it on scenes with planar surfaces.

If you are interested and need more info reach out to TAs.

Reference

@Misc{ferrer2022,
  author = {Gonzalo Ferrer},
  title = {Lectures on Perception in Robotics},
  howpublished = {\url{https://github.com/SkoltechAI/Perception-in-Robotics-course-T3-2022-Skoltech}},
  year = {2022}
}