To get updates from our aUToronto self-driving car team, please visit this page.
Dynamic Systems Lab News
Monday August 10th, 2020
In our upcoming RA-L paper, we learn robot parameters for trajectory estimation in a Gaussian variational inference setting. Our method works without ground truth, i.e. with only noisy measurements including outliers.
Find the paper here: www.dynsyslab.org/wp-content/papercite-data/pdf/wong-ral20b.pdf
Friday August 7th, 2020
Prof. Angela Schoellig presented talk about safe learning control at the IFAC 2020 Learning for Control Tutorial along side Andreas Krause, Francesco Borrelli, and Shie Mannor. Find it below!
Tutorial Webpage: www.ifac2020.org/program/tutorials/learning-for-control/
Slides: www.dynsyslab.org/wp-content/papercite-data/slides/schoellig-ifac20-slides.pdf
Friday July 17th, 2020
We present our perception-aware model predictive controller at the IFAC World Congress 2020. The controller accounts for camera orientation and location of visual landmarks to fly fast but reliably.
Paper: www.dynsyslab.org/wp-content/papercite-data/pdf/greeff-ifac20.pdf
Wednesday July 15th, 2020
Want to set up your own ultrawide-band (UWB) indoor localization system? But where to put the radio sensors? Check out our work at the IFAC World Congress 2020 to help you find the optimal sensor locations.
Paper: www.dynsyslab.org/wp-content/papercite-data/pdf/zhao-ifac20.pdf
Thursday July 9th, 2020
Throwback to last year's stage of the SAE AutoDrive Challenge! Find our systems paper summarizing aUToronto's winning self-driving car design from the competition published in the Journal of Field Robotics.
Paper: arxiv.org/abs/2004.08752
Monday July 6th, 2020
Can you localize a UAV by comparing its view with Google Earth images? We demonstrate accurate localizations of a multi-rotor UAV with a gimballed camera and an IMU using Google Earth to estimate the global pose on 7.1km of real world data with variable lighting conditions. Presented at ICRA2020
Paper: www.dynsyslab.org/wp-content/papercite-data/pdf/patel-icra20.pdf
Questions? ICRA Slack tuc07_6
Saturday July 4th, 2020
Selecting the best hyperparameters can be a difficult step in designing state estimators. We present a novel motion prior for continuous-time estimation and present how to train its hyperparameters using data at ICRA2020. In lidar localization experiments, we show a translation error improvement of up to 68% compared to previously used motion priors.
Paper: www.dynsyslab.org/wp-content/papercite-data/pdf/wong-ral20.pdf
Questions? ICRA Slack tud07_2
Thursday July 2nd, 2020
Collision-free trajectories are essential for safe swarm flight in cluttered environments. We present a robust and scalable algorithm for multi-robot motion planning at ICRA2020.
Paper: www.dynsyslab.org/wp-content/papercite-data/pdf/luis-ral20.pdf
Questions? Slack Channel mob11_1
Thursday June 25th, 2020
Safe receding horizon control in the presence of varying model error is a challenge. Our ICRA2020 paper proposes to learn a context-aware cost prediction error to achieve reliable performance in the presence of varied model accuracy.
Questions? ICRA20 Slack channel moc08_6
Paper: www.dynsyslab.org/wp-content/papercite-data/pdf/mckinnon-icra20.pdf
