We've had a very busy first day setting up our system and getting it ready for the competition next friday. There were quite some issue with our camera calibration, but we were able to fix these with another lens.
In the meantime, Fabian has been 3D-printing some extra parts to make our cable guide more robust.
Tomorrow, we hope to fully test our pipeline with the new camera set-up so we'll be ready for the stowing task next friday!
With just a few days to go until the Amazon Picking Challenge in Leipzig, team Applied Robotics is getting ready for the challenge.
We have been upgrading our camera setup and our gripper. Also, last weeks we have been spending most of our spare time on improving our vision pipeline and implementing our tote-picking pipeline.
Even during our normal day jobs, our laptops have not been idle. They have been running tests to see if our motion planner returned collision-free paths.
Tuesday, we will depart for Leipzig and the competition will start on friday. Keep tuned for regular updates!
During last week's RoboBusiness conference in Odense, Denmark, both Dutch teams exhibited their robot and prepared for the Amazon Picking Challenge by engaging in a friendly competition.
Team Applied Robotics arrived tuesday evening, the day before the conference started and we wasted no time setting up our system. Setting up our hardware was done in under 45 minutes, but our software was not yet fully tested and still needed quite some tweaking. Luckily wednesday was a relatively quiet day at the exhibition floor, so we could use our time testing the system and implementing a lot of improvements to our path planner.
Among other things, we are now able to publish collision objects and an octomap and our path planner is able to reliably plan the safest path in and out of the bins.
Friday was the day of our competition against Delft. We had agreed on a subset of items to be picked from the shelf and recruited an arbitrary judge to randomly place the items in the shelf. A large crowd had gathered to watch as the judge counted down until the start of our 10-minute match.
We were off to a quick start, and within 40 seconds we dropped our first item in the tote. Team Delft, however had some driver problems and had to opt for a restart. By the time Delft had restarted, we were already in a comfortable lead of 3 items. Picking up the rest of the items went smoothly, apart from the last one, which was placed in such a way our path planner could not find a collision-free path. Meanwhile, Delft was closing in on our score of 5 out of 6 items and deposited their third items with 2 minutes still on the clock. Luckily for us they did not manage to pick another item before the time ran out and therefore we won with 5 items against 3.
This challenge was a great opportunity for both teams to prepare for the real competition in Leipzig and we would like to thank team Delft for the great atmosphere and organization of this event.
Team Applied Robotics was challenged by Team Delft to show off our robot at the RoboBusiness Europe Conference in Odense, Denmark. (1st of june untill the 3rd of june). Of course we were thrilled to take up this challenge, also because it is the best preparation we can get for the actual Amazon Picking Challenge in July.
Our two teams will be demonstrating our robot systems during the conference in a friendly competition. This will allow us to gauge the competition and fine-tune our strategy for the competition in Leipzig.
For the new Amazon Picking Challenge we decided to improve our gripper design. The vacuum concept we used last year worked pretty well, but since this year the bins will be more crowded, our gripper needs to be even thinner.
Also for the bin-picking part of the challenge we could use an extra degree of freedom.
To make this new design, we've enlisted the help of last-year's mechanical engineer Maurice Ramaker and our latest addition to the team: Fabian Gouwens! We have developed a lot of different concepts and decided on two concepts for prototyping.
We have 3D printed the two concepts and put them to the test.
On June 30th, 2016, the second installment of the Amazon Picking Challenge will take place as part of the Robocup event in Leipzig, Germany.
We, as Team Applied Robotics, are happy to announce that we have been selected to participate again in the Amazon Picking Challenge!
One of our team members from last year, John Gardenier, has left for a PhD position in Sydney and will not participate this year.
However, we are very happy to welcome Wim Abbeloos as a new team member! He joins Berend Kupers and Simon Jansen to form this year's core development team.
We' re very excited for the challenge this year, which includes not only picking items from the shelf. but also picking items from a bin and stowing them onto the shelf.
We've restarted development and will from now on regularly update this blog in the coming months.
After participating in the inaugural Amazon Picking Challenge held alongside the IEEE ICRA 2015 conference in Seattle, we would like to describe our system in more detail. Our system was designed and implemented over a six-month period in our free time on a shoestring budget and delivered satisfactory results, resulting in a 10th place out of 28 participating teams from all over the world.
One of the first design decisions we made was to use a vacuum gripper as a mechanical solution was too large to be able to grab objects placed in the back of the shelf. An additional benefit was that compliant gripping was easy to achieve.
The robot arm we used was a UR5. Because we had access to the same robot for testing as that would be supplied at APC, we did not have to ship the actual robot. We did, however, need to ensure the robot base and gripper could be mounted and transported easily. We designed a base that could be taken apart completely and reassembled quickly. The gripper and camera bracket was also quickly mounted onto the end of the robot arm with just a couple of bolts. After assembly only two things needed to be calibrated: the position of the camera with respect to the robot arm and the position of the robot with respect to the cabinet. By defining a clear procedure for the set-up and calibration, the system was up and running within an hour.
The software architecture was based on ROS with separate modules for strategy, perception, motion planning and the interface to the UR controller. The motion planning used a heuristic approach to create Cartesian paths between the pick-up position (as determined by the perception module) and the drop-off position. For checking the Cartesian paths for collisions and reachability, URDF models of the robot and the shelf were used. After ensuring a collision-free path had been found, the robot controller executed the linear paths.
To acquire depth information, we used the Kinect V2 sensor mounted on the robot arm. This gave us sufficient resolution although issues with reflection due to the Time of Flight technique meant we needed careful filtering of the point clouds before processing.
For distinguishing objects from the shelf, recognising them and determining their poses, we took a very pragmatic approach. Because the pose of cabinet was known, we could easily filter the required bin from the pointcloud. Then we used basic PCL functionality to segment the objects. Here, we assumed that objects were not touching. Although this was true for the majority of the bins, we knew that picking all objects would be unlikely. From each object segmented from the shelf we determined the dimensions and matched them against the target object. If there was a match, the grasp poses of the object were determined by approximating the object by a bounding box. Using this approach, we were able to correctly pick approximately 80% of the objects in realistic shelf set-ups.
