- 37
- 513 532
Angela Schoellig
Приєднався 27 тра 2009
Welcome to the Dynamic Systems Lab (2018)
If you come to our lab, this is what you can see! We pulled a camera out when we welcomed visitors to the lab recently, to create this video for all of you. The video features live demonstrations of our lab as of September 2018.
The demonstrations include:
* Cooperation between a swarm of nano-quadcopters.
* Deep neural networks for impromptu trajectory tracking for a quadcopter.
* Safe predictive control for a ground robot.
* Lane detection and following for a self-driving vehicle
* Vision-based navigation solution for emergency return of UAVs when communication and GPS fails.
Work by Adam Heins, Bhavit Patel, Carlos Luis, Chris McKinnon, Felix Berkenkamp, Filip Medinac, Jingyuan Hou, Karime Pereida Pérez, Keenan Burnett, Mario Vukosavljev, Melissa Greeff, Michael Jakob, Michael Warren, Mohamed Helwa, Mohammad Nahangi, SiQi Zhou, Stephanie Marton, Thomas Bamford, Xintong Du, Zachary Kroeze, and Professor Angela P. Schoellig
Find more details: www.dynsyslab.org/research/
See also:
www.dynsyslab.org
www.schoellig.name
autodrive.utoronto.ca
Music:
High Technologic Beat Explosion - Loyalty Freak Music
Source: freemusicarchive.org/music/Loyalty_Freak_Music/ROBOT_DANCE_/Loyalty_Freak_Music_-_ROBOT_DANCE__-_02_High_Technologic_Beat_Explosion
License: creativecommons.org/publicdomain/zero/1.0/
The demonstrations include:
* Cooperation between a swarm of nano-quadcopters.
* Deep neural networks for impromptu trajectory tracking for a quadcopter.
* Safe predictive control for a ground robot.
* Lane detection and following for a self-driving vehicle
* Vision-based navigation solution for emergency return of UAVs when communication and GPS fails.
Work by Adam Heins, Bhavit Patel, Carlos Luis, Chris McKinnon, Felix Berkenkamp, Filip Medinac, Jingyuan Hou, Karime Pereida Pérez, Keenan Burnett, Mario Vukosavljev, Melissa Greeff, Michael Jakob, Michael Warren, Mohamed Helwa, Mohammad Nahangi, SiQi Zhou, Stephanie Marton, Thomas Bamford, Xintong Du, Zachary Kroeze, and Professor Angela P. Schoellig
Find more details: www.dynsyslab.org/research/
See also:
www.dynsyslab.org
www.schoellig.name
autodrive.utoronto.ca
Music:
High Technologic Beat Explosion - Loyalty Freak Music
Source: freemusicarchive.org/music/Loyalty_Freak_Music/ROBOT_DANCE_/Loyalty_Freak_Music_-_ROBOT_DANCE__-_02_High_Technologic_Beat_Explosion
License: creativecommons.org/publicdomain/zero/1.0/
Переглядів: 3 534
Відео
Sneak Peek: 2017 Summer Undergraduate Research Project
Переглядів 1,8 тис.6 років тому
A glimpse into the work of this year's summer undergraduate research team. More to come! More details about our work at: dynsyslab.org. Work by Xintong Du, Philip Huang, Qiyang Li, Quinlan Sykora, and Angela P. Schoellig at the Dynamic Systems Lab, University of Toronto.
Neural Networks for Improved, Real Time Trajectory Tracking--An Interactive Application
Переглядів 4,2 тис.7 років тому
This video shows the trajectory tracking performance improvement of a quadrotor on arbitrary hand-drawn trajectories using deep neural networks. More details about our approach can be found in the arXiv paper: arxiv.org/abs/1610.06283 Work by Qiyang Li, Jingxing Qian, Zining Zhu, Xuchan Bao, Mohamed K. Helwa, and Angela P. Schoellig at the Dynamic Systems Lab, University of Toronto, dynsyslab.o...
A Framework for Multi-Vehicle Navigation using Feedback-Based Motion Primitives
Переглядів 6817 років тому
In this video, we demonstrate a novel modular framework for safe and robust motion planning. The result is a hybrid controller that concatenates predefined motion primitives. Importantly, motion primitives are defined as feedback controllers, which achieve basic maneuvers such as moving forward, backward, left, right, and diagonally. The video shows a collection of scenarios, featuring single-v...
On the Construction of Safe Controllable Regions (CDC 2016 presentation)
Переглядів 2467 років тому
Extended version of the paper: arxiv.org/abs/1610.01243 Authors: Mohamed K. Helwa, Angela P. Schoellig Abstract: This paper studies the problem of constructing in-block controllable (IBC) regions for affine systems. That is, we are concerned with constructing regions in the state space of affine systems such that all the states in the interior of the region are mutually accessible through the r...
Short Video: Autonomous, Learning-Based Driving in Rough Terrain
Переглядів 4417 років тому
This video shows how a learning controller can compensate for unmodeled sideslopes. Details of the research are found in this publication: www.doi.org/10.1002/rob.21587. Work by C. J. Ostafew, J. Collier, A. P. Schoellig, and T. D. Barfoot, University of Toronto, Institute for Aerospace Studies. See also: www.dynsyslab.org
Innovations in Mining: Real-time Fragmentation Analysis Using UAV Technology
Переглядів 4917 років тому
In this video, we show preliminary results of using UAVs for rock fragmentation analysis. For technical details, check the corresponding paper on arXiv: arxiv.org/ftp/arxiv/papers/1607/1607.04243.pdf Work by Thomas Bamford, Prof. Kamran Esmaeili (University of Toronto, Civil Engineering), and Prof. Angela P. Schoellig (University of Toronto, Institute for Aerospace Studies). The research was co...
A Modular Framework for Motion Planning with Feedback-Based Motion Primitives
Переглядів 3 тис.7 років тому
In this video, we demonstrate a novel modular framework for safe and robust motion planning. The result is a hybrid controller that concatenates predefined motion primitives. Importantly, motion primitives are defined as feedback controllers, which achieve basic maneuvers such as moving forward, backward, left, and right. The video shows a scenario in which a quadrocopter must safely reach a ta...
Data-Driven, Force-Based Interaction with Quadrotors
Переглядів 3,5 тис.8 років тому
This video demonstrates a data-based approach for quadrotors to react to aerodynamic disturbances. Instead of estimating air flow which requires complex models and/or specialized sensors, the quadrotor estimates external forces and torques from only position and attitude measurements, and learns to associate forces with desired actions through machine learning. For details, check the correspond...
Four Flying Robots Learn Together
Переглядів 1,9 тис.8 років тому
In this video, we applied our Distributed Iterative Learning Control (ILC) approach to a team of four quadrotors. Four vehicles learn to synchronize their motion. A previous experiment with two quadrotors is found here: ua-cam.com/video/Qw598DRw6-Q/v-deo.html The goal of our work is to enable a team of robots to learn how to accurately track a desired trajectory while holding a given formation....
Distributed Iterative Learning Control for a Team of Two Quadrotors
Переглядів 1,7 тис.8 років тому
This video shows our distributed iterative learning algorithm in action for a multi-agent system consisting of two quadrotors. The goal of our work is to enable a team of robots to learn how to accurately track a desired trajectory while holding a given formation. We solve this problem in a distributed manner, where each vehicle has only access to the information of its neighbors and no central...
3-min Research Pitch on Automatic Controller Design for Complex Safety Specifications (ICRA 2016)
Переглядів 8648 років тому
by Marijan Vukosavljev, Ivo Jansen, Mireille E. Broucke, and Angela P. Schoellig Paper: www.dynsyslab.org/wp-content/papercite-data/pdf/vukosavljev-icra16.pdf Research Video: ua-cam.com/video/l4vdxdmd2xc/v-deo.html
3-min Research Pitch on Safe, Automatic Parameter Tuning for Robotics (ICRA 2016)
Переглядів 1,5 тис.8 років тому
by Felix Berkenkamp, Angela Schoellig, Andreas Krause Paper: www.dynsyslab.org/wp-content/papercite-data/pdf/berkenkamp-icra16.pdf Research Video: ua-cam.com/video/GiqNQdzc5TI/v-deo.html
Safe Controller Optimization for Quadrotors
Переглядів 8 тис.8 років тому
This video demonstrates a new approach for safely optimizing controller parameters on a real system, without failures or crashes. The approach modifies a recently developed safe Bayesian optimization algorithm and applies it to a real system for the first time! More details about our approach can be found in the arXiv paper: arxiv.org/abs/1509.01066 The source code of the algorithm is openly av...
Safe and Robust Quadrotor Maneuvers Based on Reach Control
Переглядів 2,5 тис.9 років тому
In this video, we demonstrate a novel control concept for safe and robust quadrocopter flight. We formulate the problem as a reach control problem. The result is a hybrid controller with different affine controllers being active in different parts of the state space. The video shows the very first demonstration of reach control in experiments; reach control had so far only been developed in the...
Good day. Do you have an e-mail address? How can I write to you?
which parameter you considered for training in GP. for example the (Inputs and Targets).
Angela Schoellig, hi, I read your paper "Safe and Robust Learning Control with Gaussian Process", it's really interesting and I have some questions., I want to know how do you split the priori model f(x,u); and unknown model g(x,u); with the system states and inputs x & u. and also which parameter you considered for training in GP. please.
Machine Learning in Robotics as a topic for my Bachelor Project was suggested to me by a Professor here in Germany. I hope it will be the right choice for my future work
It's a wonderful study. I am interested in that research. But how do you perform matrix estimation of the drone's equation of state using Gaussian process regression? I want to know more. Could you tell me if you like? Or could you attach the files (model, code, etc.) used for matrix estimation?(mizutani407@gmail.com)
Great job.
I am so dedicated to work in this lab
This is superb 😍
What are item that you use of your project
This is really cool. I wants to move my research to this. did you try this outside to see whether it can keep good balance with strong wind or not?
This is amazing! does this method work in cases that we don't know where the obstacles are?
Impressive, I am one of many Udacity FCND students progressing through the course materials right now and learning a great deal from your instructions. Thanks Angela!
I feel like clapping!
Good job.
Nice! It never over shoots (almost) which is a very hard thing to achieve. I have done something similar: ua-cam.com/video/jWs9Ur9NiRo/v-deo.html I would love to see links to relevant publications in the video description.
Hi.that is fantastic. But what kind of algorithms do you use?
this job inspired me to work with quadrotors, thank you very much.
My like is video
Me encanta tu video
Is there a conection betwenn machine learning and control systems? I am really interested in control systems as a whole but also like the math and the coding for training neural networks and dealing with machine learning in Matlab for example. What would you recommend? This video is awesome by the way! Thanks for posting it
Santiago Infantino Moreno In my opinion, there is a nice connection between machine learning and control. Machine learning is always useful, and useful in control, when it is difficult or impossible to describe the system behavior in terms of mathematical equations based on our prior knowledge (e.g., based on Newton's equations).
Awesome! Great work guys!
The minute 1:07 is very controversial, is not the technology in itself that makes this society a better society, but the use that you do of it.
Amazing!!!!
Very interesting work Dr. Schoellig. I am currently working on human robot interaction research; however I am very interested in highly dynamical systems such as quad copters and bipedal robots. I would definitely want to work on a similar project as yours in the near future.
I found a -74% coupon to Learn Matlab Udemy course www.udemy.com/learn-matlab/?couponCode=ML25
Great work! My question is the GP already takes into account the uncertainty, so what is the benefit of robust framework? Just stability? I think both stability and uncertainty can be considered in GP model. Would you please give some clarification in this regard?
Ali Baheri The GP learns a model with uncertainty bounds (i.e. it learns the model behavior and also gives us information about how certain it is about what it learned). What do we do with this model then? We need a control framework that guarantees stability for this uncertain model, this is robust control
Pretty cool, but you would need to do this after every hardware configuration change. or if there was the slightest hardware problem. Maybe it could be done by using an internal simulator of the weight distribution, gravity and physics of the air as to calculate its moves backwards from it target position to its current position.
I have been interested in ILC, but unable to get a good head start or lead. I am in my year 3 now and I have done some control system courses and interested in doing research in Mobile Robots with application in Aerial robots. What are the list of mathematical background i need to get started as I find it a bit difficult to understand some of the mathematical aspect of it when reading journals and books.
This is interesting and valuable research that could truly benefit the future of our planet. Where can I learn more?
Seriously??
New video of my group! First proof-of-concept demonstration of reach control on quadcopters for safe & robust flight!
das ist super
Invigorating !
This is amazing work, Angela! Congratulations! :D
Vou comentar no meu idioma agora, parabéns pelo trabalho! Desejo muito sucesso aos teus alunos.
When I see these, I remember The Invincible by Lem...
Excelente presentación, dentro de poco estarán invadiendo el espacio aéreo con estos equipos
Really inspiring, keep working Angela
Good grief, that's the most stable quad platform i have ever seen. It had practially zero wobble at all. Your sensing resolution & control algorithms must be epic.
Armageddon? That doesn't sound good...
Proud of my summer students:@ardrone and ROS | Quadrocopter Bounce & Tag Following @UofT
Nice, did they learn the celebratory backflip maneuver too? :)
We're waiting for the "Harlem Shake" version!! :D
A computer generates and sends the trajectory commands to the vehicle. Was this your question?
No, we use model-based learning techniques. Refer to our webpage for details.
!!YAY!!!!
Go home non-learning quadrocopter, you are drunk.
Amazing work...
Is just bizarre that my computer be capable of fly my RC model aircraft better than me!! Amazing work...
They are searching for John Connor