What’s New in Robotics? 07.07.2023
News roundup for this week look at RoboCat and the start of the age of all-purpose robots, then the 10x productivity increase of Lockheed’s “smart” drilling robot, then here’s why robots need better batteries, followed by the world’s trash-picking champ with 2700 picks per hours, and lastly, Comau also dives into bin-picking draws with its super-fast cell picking.
This is the era of all-purpose robots
Trying to get robots and cobots to quickly switch between multiple tasks was, until now, a tricky task to pull off. Most robots today have a specific purpose, in that they do one task well until needed to do another, at which point they require more programming.
General-purpose robots, with the ability to easily automate almost any given task when needed, have long been elusive. However, the new Google RoboCat may be a breakthrough answer, offering a new level of intelligence, plus flexible automation, increased productivity and lowered costs. In short, a manufacturer’s dream tool.
Created by Google based in the UK DeepMind subsidiary, RoboCat is a self-developed AI agency that “learns to operate different robotic arms — and perform different tasks — with just 100 examples, and can generate its own training data to help sharpen performance,” says DeepMind. The spin-off agent would then have the robot practice the task repeatedly, physically and virtually — on average, 10,000 times.
The AI behind RoboCat is called Gato, which can process actions, images and language in both the real world and simulated environments. DeepMind feeds Gato a training set filled with robotic arms that perform hundreds of different tasks.
Ultimately, RoboCat’s library of general-purpose robotic arm movements upgrades itself to “data set of millions of tracks, from both real and simulated robotic arms, including self-generated data.” RoboCat can learn to operate a new robotic arm in a few hours.
“After observing 1,000 human-controlled demonstrations, collected in just a few hours, the RoboCat was able to maneuver the new arm deftly enough to take gears successfully 86% of the time.”
Not perfect, but pretty good!
The real world of “smart” robots.
What would a robot or cobot with intelligence look like in the real world? They are known by their deeds. Preliminary results show “smart” robots being incredibly powerful in manufacturing, such as a 10-fold increase in productivity Lockheed Martin just experienced.
Lockheed Martin identified a use case for implementing AI focused on the typical robotic work cell used by its manufacturing cobots: drilling holes.
Can AI “empower commercial cobots with greater intelligence and understanding of his body and the task at hand while ensuring the required qualities and tolerances are achieved?”
Xaba’s xCognition “synthetic brain”, which specializes in manufacturing, was used in production tests. Xaba claims that it is the first AI-driven robotics developer and CNC machine controller. Unlike analytical AI which is used for predictive modeling, Xaba says that Industrial AI focuses on the physical world. It can model and drive vehicles, complex machining centers, and more complex industrial applications.
The test consists of two phases:
Phase 1: Assess the cobot’s performance in maintaining accurate and consistent track positioning with and without Xaba’s xCognition.
Phase 2: Perform a series of drilling tests on an aluminum test plate with a certain position tolerance.
Results: Based on the data collected, xCognition improves the accuracy and consistency of commercial cobots by a factor of 10-fold.
Matthew Galla of Lockheed Martin, an aeronautical applications engineer, commented: “Testing with the xCognition controller allowed us to rethink how we could accelerate innovation in manufacturing.”
From the outset, as witnessed in Lockheed’s testing, the impact of AI in aiding the creation of robots and cobots seemed staggering.
Better battery for robots
Fire at the Ocado warehousestarted with an automated moving robot (AMR) that recharged its batteries, burned for three days, lost 300,000 customer orders and 1,100 robots in the process for a total loss of $48 million.
The warehouse burned to the ground in a blaze three days after a fault in a battery charging unit burned the plastic lid of one of the grocery picking robots.
Some reports suggest that the robots were still moving when the firefighters arrived, hindering their ability to control the blaze.
It was not an isolated incident. Another battery-related fire started when three warehouse robots collided. Once turned on, the robot’s lithium-ion battery is very difficult to extinguish. As AMR proliferates at the accelerated pace of automating warehouses, the possibility of more fire hazards becoming more likely.
And warehouse fires aren’t the only problem plaguing the robot’s lithium-ion battery, nor are warehouses the only trouble spot.
A recent article in the journal Nature, Robots Need Better Batteries, to quote a mobile robot — on wheels, tread, tracks, or on two or four legs — can’t tolerate very large batteries. Boston Dynamics sells a four-legged dog-sized robot named Spot that weighs about 32 kg — one-eighth of which is the battery, which needs to be recharged after just 90 minutes.
“The energy density is still quite far from the power we need for robotics,” said Ravinder Dahiya, an electrical engineer specializing in robotics at Boston’s Northeastern University.
Looking ahead, lithium-ion technology is aging rapidly, with “less and less room for improvement,” says Richard Schmuch, chemist at Germany’s Fraunhofer Research Institute for Battery Cell Production. “Lithium itself is rare and expensive. The same goes for cobalt, another important element that can make up 20% of the cathode weight in lithium-ion batteries for electric vehicles.”
Like lithium, sodium is an alkali metal, and the chemical properties of the two are so similar that researchers are pursuing sodium-ion batteries as a way to solve problems with lithium ions. Sodium in seawater and salt deposits is easier to harvest, which eliminates the supply chain problems that arise from the expense and scarcity of lithium.
China Contemporary Amperex Technology (CATL) in Ningde — the world’s leading manufacturer of lithium-ion batteries for electric vehicles — launched the first generation of sodium-ion batteries; and this April, Cherry’s car in Wuhu, China, announced plans to install CATL sodium-ion batteries in its cars.
Sodium-ion battery advocates say it offers the best prospects for preventing the potentially lethal fires seen with lithium batteries, as well as for increasing energy density and reducing costs.
Better batteries for robots may be around the corner.
Blitz picking at 2700 per hour
lardThat German vacuum gripper developers, must have been a little tired of seeing trash picking cells operating at pickup speeds less than considered appropriate. So, it developed itself, and in the process, set a new speed record for trash pickup. Just say it flash capture with 2700 shots per hour!
Schmalz blitz chose the best closest rival Pickit (2100 per hour) by 600. Both far exceed the pick-and-place average robot speed 300 per hour.
With the average number of items in an ecommerce order just over three, champ picked Schmalz outperformed Pickit with 200 orders per hour.
In collaboration with German integrator Körber, Schmalz built his own fruit picking cell using a pair of SCARA robots (Körber built a conveyor system).
“Robots work simultaneously and can change their gripper in a second, depending on whether they’re handling boxes, vials, or bags. Two 3D cameras check the box after each shot.
“The software analyzes this input, calculates and controls the next grip. Multiple grip algorithms operate in parallel to achieve the best retrieval performance. This system uses a five-axis gripper that can reach all corners of the crate.”
Schmalz developed a pick accelerator “kit” that optimizes the interaction between the gripper, robot, camera and pick software called the Schmalz Solution Kit.
Comau joins the trash scavenger hunt
Milan-based developer of automation and robots Coma has also joined the trash picking contest with hers MI. RA voterswhich uses two high-resolution laser sensors and a central camera, combined with virtual simulation tools and predictive algorithms, to optimize lane management and achieve a collision-free trajectory.
Interestingly, the MI.RA Picker can be used with any build robot, bin or gripper to recognize, find and capture up to 40 randomly placed items every minute, which translates to up to 2400 picks per hour. That’s the Pickit’s better speed, but still falls short of Schmalz’s industry-leading pickup kit of 2,700 takes per hour.
