This weekend we drove over to the lovely University of Michigan campus to attend the workshops at RSS 2016. During the daytime, we listened to several heated exchanges about the future of hyper-parametric function approximation (i.e. Deep Learning) in robotics. During the night, we spent some time arguing over the best local Michigan brews. A great time all-around!
Lee and Valentin at RSS 2016.
The Deep Learning workshop at RSS 2016.
We visited MDA in Brampton, Ontario today for the 2016 Space Vision and Advanced Robotics Workshop. Jacob presented his work on entropy-based extrinsic sensor calibration, while Valentin talked about first responder localization using foot-mounted IMUs. We took home Second Prize (space pens!) in MDA’s Best Demo competition!
“Infrastructure-free First Responder Localization” won Second Prize in the Best Demo competition!
“Extrinsic Calibration of 2D Lidars Through Entropy Minimization”
Valentin presented his work on PROBE-GK: Predictive Robust Estimation using Generalized Kernels (in collaboration with the MIT RRG including Will Vega-Brown) at ICRA 2016 in Stockholm. A great turn out for the 3-minute spotlight talk, and lots of great questions at the interactive poster!
Valentin and Will (from Nick Roy’s group at MIT) at ICRA 2016.
Valentin presenting in the Visual-Based Navigation session.
We are grateful to NVIDIA for providing the lab with a state-of-the-art Titan X GPU. Looking forward to some cutting edge research on deep learning!
Congratulations to Valentin, who will be presenting “PROBE-GK: Predictive Robust Estimation using Gaussian Kernels” at ICRA 2016 in Stockholm, Sweden! See you there!
Congratulations to Lee and Valentin on receiving prestigious NSERC scholarships from the Government of Canada. Lee received the PGS-D3 and Valentin received the CGS-D3. Both excellent results!
Congratulations to Valentin and Lingzhu on their well-attended presentations at the International Conference on Intelligent Robots and Systems 2015 in Hamburg!
Lingzhu’s presentation on robust SLAM.
Valentin’s IROS presentation on PROBE.
Valentin Peretroukhin will present his work on predictive noise modelling for visual-inertial navigation at IROS 2015 in Hamburg, Germany. This research enables learning of robust sensor noise models – making it easier to build consistent estimators. Congratulations Valentin!
We have two papers accepted at the 2015 Conference on Computer and Robot Vision in Halifax, Nova Scotia. Congratulations, Lee, Valentin, Jacob and Faizan!
Lee Clement’s work on a monocular Visual Teach and Repeat (VT&R) pipeline for mobile robots will be presented at the FSR 2015 conference in Toronto in June. The paper demonstrates that basic information about camera placement can be employed in a monocular VT&R system to enable kilometre-scale route repetition with performance nearly identical to that of stereo VT&R. Congratulations Lee!
Example of noise model used in Lee’s paper to enable robust navigation.