Today, I’m excited to welcome Xometry, a leading marketplace for on-demand manufacturing as a quarterly sponsor of Weekly Robotics! I’m incredibly excited about this partnership because Xometry actively works with the robotics industry, among others. As usual, the publication of the week section is manned by Rodrigo. Last week’s most clicked link was the ROS 2 Visual Programming Tool, with 14.7% opens.
Weekly Robotics is being developed thanks to the Patreon supporters and the following business sponsors:
Plan a fulfilling career that does more good with 80,000 Hours
80,000 Hours is a nonprofit dedicated to helping people find careers that help solve the world’s most pressing problems — from pandemics to artificial intelligence and nuclear war. Read in-depth, evidence-based profiles of high-impact careers, like AI safety technical research, and find relevant vacancies on the 80,000 Hours job board.
RT-2: Vision-Language-Action Models
“We study how vision-language models trained on Internet-scale data can be incorporated directly into end-to-end robotic control to boost generalization and enable emergent semantic reasoning. Our goal is to enable a single end-to-end trained model to both learn to map robot observations to actions and enjoy the benefits of large-scale pretraining on language and vision-language data from the web”.
Check out the examples on the page to better understand the capabilities of this system, such as “Push the ketchup to the blue cube” or “Move banana to Germany”.
Robot Raconteur is a framework for distributed control and component interfaces. Looking at the quick start code, the approach seems quite different than ROS - it seems to fetch an object through a TCP connection. You can learn more details about how it works by checking out the documentation.
Understanding Battery Performance of IoT Devices
Batteries are hard. In this article, Tyler Hoffman looks into Li-Ion batteries from the perspective of a hardware engineer. The report is full of excellent advice, so if you need to incorporate a battery in your design, especially if the battery state is communicated to the user, I highly recommend taking a look.
Kappe is a tool for converting and splitting MCAP files. Looking at the readme file, the operations I find useful given my experience to date, are topic renaming, removing topics, changing timestamps, rotating point clouds, modifying TFs, and trimming. I can’t wait to add Kappe to my toolbox. For more information and the discussion about the project, check out ROS Discourse
Bot Inspired by Baby Turtles Can Swim Under the Sand
Roboticists at the University of California San Diego created a robot that can ‘swim’ in granular environments such as sand. The robot can move with a speed of 1.2 mm per second (4 meters per hour) and even has fins! For more information about this work, check out the open-access paper Toward Robotic Sensing and Swimming in Granular Environments using Underactuated Appendages.
Embrace Complexity; Tighten Your Feedback Loops
You are reading a robotics newsletter, and you see a feature mentioning feedback loops, and you might be thinking you are about to learn some control theory. Then you open it up and see a slide saying: “This is all going to hell anyway”, and you are now even more sure you are in control systems territory, but no. This piece is about organizational feedback loops and touches on working with people, leading large projects, and scaling organizations. I found the article to be thought-provoking and decided to feature it here.
Mapping The Depths With An Autonomous Solar Boat
Here is an autonomous solar boat built from PVC pipes used as pontoons. The vehicle uses differential thrust for steering and a Ping echo sounder for Blue Robotics. The robot can capture interesting depth maps of ponds.
Publication of the Week - CBGL: Fast Monte Carlo Passive Global Localisation of 2D LIDAR Sensor (2023)
Most robots without a global positioning system will depend upon an initial guess to localize on a map. This paper presents a method for passive global localization using 2D LiDAR sensors. The author takes the strengths of usual methods by dispensing any environmental structure information and adapting to different sensor angular ranges. The tests showed better results against efficiency and precision when compared to state-of-the-art methods. You can check the code for ROS in C++ in this GitHub repository.
Waymo pumps the brakes on self-driving trucks
“Waymo has announced that the company is focusing its efforts and investments on its autonomous ride-hailing service and pushing back the timeline of its commercial and operational trucking efforts”.
AutoStore to pay Ocado $256M in patent lawsuit settlement
“Ocado Group and AutoStore Holdings have announced a complete settlement of all claims between the companies in their global patent dispute. The dispute began in October 2020, when AutoStore asserted that Ocado infringed upon six of its patents, and included lawsuits in both the UK and the US”.
Astrobotic earns $34.6M NASA partnership for lunar power demo
“Astrobotic has won a $34.6 million NASA Tipping Point partnership to demonstrate the power transmission abilities of its LunaGrid-Lite on the lunar surface. This will be the first-ever transmission of high-voltage power across the lunar surface.
LunaGrid-Lite will transmit power from a lunar lander to a tethered rover. During the demo, an Astrobotic 6U CubeRover will unreel 1 kilometer of cable from an Astrobotic lunar lander and receive the high voltage power. The demonstration will serve as a pilot for Astrobotic’s LunaGrid service, which aims to deliver commercial power service by the Watt to future customers on the lunar surface”.