Weekly Robotics #14

1) Fast line-following robots part 1: control.
https://www.a1k0n.net/2018/11/13/fast-line-following.html
INFO: In this blog post Andy Sloane goes over evolution of his line following robot for DIYRobocars meetup in SF. The article covers the evolution of a line follow algorithm that starts as a purely proportional controller and finally takes a form of a PD controller keeping a target velocity on a curve with a lookahead control. We recommend playing with the included algorithm visualizations and controller tuning.

2) Unexpected AI.
https://docs.google.com/spreadsheets/u/1/d/e/2PACX-1vRPiprOaC3HsCf5Tuum8bRfzYUiKLRqJmbOoC-32JorNdfyTiRRsR7Ea5eWtvsWzuxo8bjOxCG84dAg/pubhtml
INFO: This Google sheet is a compiled list of cases where an AI system uses the imposed rules in ways unexpected by the authors. Some of the entries are robotics related: “a robot arm with a purposely disabled gripper found a way to hit the box in a way that would force the gripper open”.

3) Cybathlon 2020.
http://www.cybathlon.ethz.ch/cybathlon-2020.html
INFO: Via linked website: The Cybathlon is a unique championship in which people with physical disabilities compete against each other to complete everyday tasks using state-of-the-art technical assistance systems. The teams have time until November 2019 to register, provide safety documentation and complete a pilot form. This video shows a trailer for 2016 edition.

4) ROSCon and IROS 2019.
https://roscon.ros.org/2019/
INFO: ROSCon 2019 has been announced and will take place from October 31st to November 1st in Macau. IROS will take place in Macau as well from November 3rd to November 8th.

5) How multi-beam flash lidar works.
https://www.ouster.io/blog-posts/2018/11/8/how-multi-beam-flash-lidar-works
INFO: The linked document is a technical note on Ouster lidar. Interestingly Ouster lidar is working with light on 850 nm wavelength (for comparison Velodyne HDL-64E is using 905nm light). The article also mentions Outster’s approach for Flash lidar and technological improvements that matter to their technology.

6) RocketLab’s satellite launch.
https://www.nytimes.com/2018/11/10/science/rocket-lab-launch.html
INFO: RocketLab’s Electron rocket just carried out a successful mission, carrying small commercial satellites to the orbit. Compared to most of commercial rockets we usually hear about the Electron is quite small. It’s height is only 17 m (56 feet) and it can carry about 220 kg (550 pounds) of payload.

7) Publication of the week - UMBmark - A Method for Measuring, Comparing, and Correcting Dead-reckoning Errors in Mobile Robots(1994).
http://www-personal.umich.edu/~johannb/Papers/umbmark.pdf
INFO: UMBMark (University of Michigan Benchmark) is a method for measuring dead-reckoning errors in mobile robots. To successfully perform UMBMark you will need to run the robot on a 4x4m path, while collecting robot position information. On page 15 of the linked document you will find a one-page summary of the procedure. If you are looking for a read that’s shorter than 40 pages, a solder spot article is quite succinct and contains a C source code with UMB results implementation.

0) Introducing the sponsored section.
https://weeklyrobotics.com/About
INFO: It has been 100 days since the launch of Weekly Robotics. Since compiling each issue takes anywhere between 3 to 6 hours we think that the sponsored section should help cover the personal costs and keep this newsletter going. Please get in touch if you would like to advertise your robotics related products or services with us.

1) Handling dead nodes in ROS
https://msadowski.github.io/ros-nodealive/
INFO: In this article the author of this newsletter explores methods of determining a state of a ROS node (especially capturing whether it’s still alive).

Careers

0) Would you like to advertise an open position in a robotics related company?
https://weeklyrobotics.com/About
INFO: If you would like us to include your open position in the hiring section please feel free to send us an e-mail.

1) Robotics & Perception Group at ETH (Zurich, Switzerland) - PhD/Postdoc in Deep/Reinforcement Learning for Computer Vision.
http://rpg.ifi.uzh.ch/positions.html
INFO: Robot Perception Group’s mission is to develop autonomous machines that can navigate all by themselves using only onboard cameras, without relying on external infrastructure, such as GPS or motion capture systems.

2) Ouster (San Francisco, CA, US) - Various Positions.
https://www.ouster.io/careers/
INFO: Ouster is developing lidar sensors (see the 5th entry of this Weekly Robotics issue).