Welcome to the repository of two subjects taught at ITMO University for Robotics and Artificial Intelligence students.
Instructor: Ivan Borisov, PhD, [email protected]
Assistant: Egor Rakshin, MSc, [email protected]
The design process is often not formalized and mostly relies on the designer’s imagination, engineering intuition, and experience. Within this subject we utilize computational design paradigm, that treats the design process as an optimization task, to optimize mechanics and control for robotic and mechatronic systems. Computational Design consists of two tasks: the user-driven the forward exploration to compute performance and behavior of a system with a given set of parameters and the optimization-driven inverse task to determine values that lead to a desired motion or behavior. The module discuss how computational design approach can be used for engineering purposes.
- Lecture notes
- Book: Hurbans, Rishal. "Artificial Intelligence Algorithms." Hurbans Rishal–Shelter Island, NY 11964 (2021) Link.
- Video course: Introduction to Computational Thinking
- Video course: Introduction to Computational Thinking and Data Science
- Papers: Computational Robotics Lab
- Papers: Disney Research
Modern robotic and mechatronic systems are complex in terms of all domains: mechanics, sensors, actuation, control, etc. To study the behavior and performance of an existing robotic system or a proposed one, we use models to focus on the essential features while keeping a reasonable tradeoff between realism and simplicity. The act of building a model is called modeling, while the process of using a model to study the behavior and performance of an actual or theoretical system is called simulation.
This module discusses modern techniques for creating and using models to study the behavior and performance of robotic systems. The module consists of lecture and practice parts. The lectures mostly give theoretical inputs on modeling such as screw theory & Lie groups to describe motion, bond-graphs to describe the interconnection of different physical systems, control strategies to steer the systems, and optimization procedures for mechanics and control. The practice part focuses on simulation skills using MuJoCo and Pinocchio.
After the module the student:
- Understands trade-offs when modeling dynamical systems
- Understands advanced Screw Theory & Lie Groups and Bond graph modeling concepts
- Understands how to use simulation to gain insight into, analyze, and optimize models
- Is able to use MuJoCo for multibody simulation
- Lecture presentations
- "Modern Robotics: Mechanics, Planning, and Control," Kevin M. Lynch and Frank C. Park, Cambridge University Press, 2017
- R. M. Murray, Z. Li, S. S. Sastry, and S. S. Sastry, A mathematical introduction to robotic manipulation. CRC press, 1994.
- Featherstone, Roy. Rigid body dynamics algorithms. Springer, 2014.
First, you need to install conda package manager. Use one of the suggested instructions depending on your OS.
After installing conda, set up the virtual environment by running the following command in the terminal
conda env create -f environment.ymlNext, activate the environment:
conda activate simrobsMore information about virtual environments you can find here.