- Vision Aided Navigation (086761)
- Autonomous Navigation and Perception (086762)
- Planning Under Uncertainty in AI and Robotics (0970252)
Vision Aided Navigation (086761)
The course focuses on fundamental topics in vision aided navigation (VAN) and simultaneous localization and mapping (SLAM), which are essential for autonomous operation in unknown, uncertain or dynamically changing environments.
Topics to be covered include: Bayesian inference, state of the art SLAM and VAN approaches, and bundle adjustment. Depending on progress, some of the following advanced topics will be also briefly covered: multi-robot cooperative localization and mapping, active SLAM and belief space planning, intro/overview of recent deep learning approaches.
We encourage student participation from multiple faculties/departments at the Technion.
- The course syllabus and more information can be found here.
- The course is managed via Piazza.
- Let me know (by email) if you do not have a Technion email.
Course tentative schedule (topics and schedule often change from one semester to another):
| Lecture week | Topic |
|---|---|
| 1 | Introduction, 3D rigid transformations, 6 DOF Poses |
| 2 | Probability basics, Bayesian inference, Extended Kalman filter |
| 3 | Projective camera geometry, Feature detection and matching |
| 4 | Structure from Motion I, Multiple view geometry, Bundle adjustment |
| 5 | SLAM and VAN |
| 6 | Graphical Models, iSAM |
| 7 | iSAM, visual-inertial SLAM |
| 8 | Active SLAM, Belief space planning |
| 9 | Midterm exam |
| 10 | Cooperative navigation and SLAM I |
| 11 | Cooperative navigation and SLAM II |
| 12 | Project presentations |
| 13 | Project presentations |
Autonomous Navigation and Perception (086762)
The course focuses on fundamental topics in planning under uncertainty (belief space planning) in the context of autonomous navigation and perception, considering online autonomous operation in unknown, uncertain or dynamically changing environments.
Topics to be covered include: Probabilistic inference, MDP and POMDP formulation, belief space planning (BSP), information-theoretic costs, search- and sampling-based planning, application to autonomous navigation and active SLAM, Gaussian processes, informative planning and active perception, an overview of (deep) learning-based approaches.
We encourage student participation from multiple faculties/departments at the Technion.
- The course syllabus and more information can be found here.
- The course is managed via Piazza.
- Let me know (by email) if you do not have a Technion email.
Course tentative schedule (topics and schedule often change from one semester to another):
| Lecture week | Topic |
|---|---|
| 1 | Introduction, Probabilistic inference, Environment representations |
| 2 | MDP & POMDP formulation, Belief space planning problem |
| 3 | Belief space planning problem |
| 4 | Search-based and Sampling-based planning |
| 5 | Search-based and Sampling-based planning |
| 6 | Information-theoretic costs/rewards |
| 7 | Application to autonomous navigation and active SLAM |
| 8 | Midterm exam |
| 9 | Informative planning |
| 10 | MDP and POMDP revisited, data-driven approaches |
| 11 | MDP and POMDP revisited, data-driven approaches |
| 12 | Project presentations |
| 13 | Project presentations |
Planning Under Uncertainty in AI and Robotics (0970252)
The course provides mathematical tools and approaches for solving planning under uncertainty problems in partially observable domains in the context of AI and robotics.
Topics to be covered include: Probability space, Probabilistic inference, MDP and POMDP problem formulations, belief space planning, information-theoretic costs, nonparametric inference, offline MDP and POMDP approaches (dynamic programming, value iteration, policy iteration, Alpha vectors), online MDP (forward search, branch & bound, sparse sampling, Monte Carlo tree search), online POMDP approaches (POMCP, POMCPOW, PFT-DPW), robust and risk-averse planning under uncertainty.