Trains are Robots, Too

Congrats to Tyler, whose paper on lidar-based train localization won best robotics paper at the conference on Computer and Robot Vision (CRV) in Vancouver this year!

Sudbury Bound

It's been two years since we were on a big field test but we're off to live in a gravel pit in Sudbury, Ontario for two weeks, once again.  We'll be testing a new version of Visual Teach & Repeat (and plan to drive 100 km autonomously), a cliff-descending tethered robot for geological mapping, and some small fixed-wing planes for aerial surveillance.  Image on the right shows the big pile of gear we're starting to gather for our trip.  We've even got the floorplan of our office truck mocked up and have been making sure our power system will work.  Hoping for no bugs, both software and of the insect variety.

Full STEAM Ahead at IROS 2015

Sean Anderson presented our paper, "Full STEAM Ahead: Exactly Sparse Gaussian Process Regression for Batch Continuous-Time Trajectory Estimation on SE(3)", today at IROS 2015 in Hamburg.  He got a huge audience and most of them are even looking at the screen!  Time to relax and enjoy the conference.

FSR 2015 is Done

Last week we hosted Field and Service Robotics (FSR) 2015 at the University of Toronto and it was a big success.  The weather was beautiful for our boat cruise and live demos and all the social events were great.   UTIAS presented 4 papers and they all seemed to be well received.  Thanks to everybody who participated in the organization and execution of the conference; it was absolutely a team effort and the results were worth it.  Just need to take care of some closing tasks and it will be done done.  There's a great article on FSR at Robohub.

Grizzlies Belong in the Woods

Spring has sprung and our Grizzly is on the prowl again.  The image on the right shows the Grizzly autonomously repeating a route using vision at 3.5 m/s through a wooded area.  This is easier said than done as there are a multitude of hazards that must be avoided.  We demonstrate the route we want to drive to the robot.  Then, our path-tracking controller gradually increases speed in sections of the path where it can keep errors within user-defined bounds; it therefore learns to drive quickly while keeping path-tracking errors very low.

Winter Wonderland

Winter is here...brrr.  We've been working on getting our visual teach and repeat navigation framework to work in some difficult conditions.  Snow is one of those conditions.  It's terrible for both traction (lots of wheel slip) as well as vision systems (very few stable features and changing geometry from one day to the next).  Multiple cameras seems to help quite a bit and we hope to have some results of this testing submitted to the next edition of Field and Service Robotics, which will be held in Toronto in June, 2015.

T-ReX Tensions Rise

T-ReX tension calibration

We're developing a new version of our climbing robot called T-ReX (Tethered Robot eXplorer).  The plan is to deploy it in a canyon in a newly discovered crater on Victoria Island in the Arctic in the summer of 2015.  The mission will be to teleoperate the robot and build a 3D model of the stratigraphy of the crater that has been exposed in the canyon.  We've completed the mechanical design and here are calibrating the tension sensor so that we can start to implement tension-based control for ascent/descent on near-vertical surfaces.   We'll be deploying the robot on the exterior of our MarsDome to emulate the canyon wall in the near future.