pre-postAI-robots
weekend-edition150

Brainy Logistics: Intelligent Robots Step Up, Part 2

Like their human counterparts, pre-AI robots may well be facing an uncertain future of huge layoffs.

Pre-AI vs. Post-AI
Sooner than later, maybe much sooner than later, every robot manufactured, regardless of make, model or use, will carry a sticker, prominently displayed, that reads either “Pre-AI” or “Post-AI”.

Like automobiles without GPS, no one is going to want a “pre-AI” robot. Across the global robotics ecosystem, it seems like every startup and newbie developer is rolling out a robot or logistics tech souped up with artificial intelligence /machine learning (AI/ML); and every established robot brand is rushing about to develop or acquire AI tech to build into or graft onto machines in their existing product lineup.

And it’s all happening fast! AI/ML-powered logistics automation was ramping up well before COVID, but most certainly COVID has increased the pressure on warehouses and distribution centers.

Why the need for speed?
According to Prologis, the San Francisco-based real estate investment trust, e-commerce requires more than three times the logistics space of brick and mortar stores. Prologis vice president Melinda McLaughlin: “Retailers generally offer a much wider product variety online, compared to in-store, and need to accommodate greater volatility in purchase activity, both of which drive the need to hold more inventory.”

Hence, gargantuan warehouses and distribution centers.

Additionally, she says, “E fulfillment also incorporates individual product picking and space-intensive parcel shipping operations. Finally, many of these spaces need to accommodate returns and the accompanying reverse logistics, and often value-add operations such as assembly. In some cases, the movement of goods can require a new warehouse that is larger in size and height.”

Add in COVID, with millions of more people ordering millions of more SKUs online, and e-commerce gets accelerated; this year up more than 25 percent in April, ahead of the 15 percent rate at the end of 2019. Top off the mix with one-day or even same-day delivery, and things could get mighty busy in the world of logistics; which they have.

All of which makes for perfect timing for an AI/ML solution to the logistics dilemma. A tailor-made opportunity for lots of bright people with inventive ideas, stacks of investor cash, and ready logistics tech to launch into e commerce fulfillment problem.

Some of the brainiacs
Dexterity, Inc. (Redwood City, CA) is one the latest in a spate of AI/ML robot developers to make an appearance. Dexterity just emerged from stealth (founded 2017) with a robust, AI-powered logistics robot…plus a neat $52 million in funding. Since Dexterity’s tech is a hardware-agnostic, robotic system add-on, look for an established robot manufacturer to take a close look at the newbie as an acquisition’s target. With Kawasaki Heavy Industries already a customer, that could well make for an interesting marriage.

The Bloomberg headline says it all: Dexterity, Inc. Introduces Intelligent Robots for Warehouse Automation that Pick, Move, Pack and Collaborate.

What’s left to do in a warehouse? Not much.

Dexterity’s robot skills were specifically designed with warehouses in mind. The company’s founder, CEO Samir Menon, said: “While robots are the backbone of manufacturing, they have historically lacked the ability to adapt and operate in dynamic environments like warehouses.”

Dexterity’s robots, reads the company’s press release “can move, and pack items using the sense of touch, and work collaboratively with one another. For instance, two robots can collaborate to pick trays or crates, and even collaboratively move them across the work-area if required. To support high performance and adaptability with safe human-robot interaction, its robots have capabilities like touch perception, computer vision, force control and contextual awareness.”

Here again is that quest for the use of arms and hands that Covariant.ai’s (formerly Embodied Intelligence), Pieter Abbeel spoke about in Part 1 of this series: Brainy Logistics: Intelligent Robots Step Up, Part 1.

Things done with legs, he says, have a decade of tech behind them, and that the leg category in logistics automation has been well-addressed by many. “The pressure now is on the hand part,” says Abbeel. “It’s about how to be more efficient with things that are done in warehouses with human hands.”

Abbeel has been all about that very task for a while now. Back in 2013, his UC Berkeley Robot Learning Lab (RLL) was teaching robots to tie a knot. Sounds simple; it isn’t. Four years later (2017), having picked up a few brain-hand-eye robot skills along the way, he and his partners, Rocky Duan, Tianhao Zhang, Pieter Abbeel, and Peter Chen (pictured), plus $7 million in seed funds, co-founded

The covariant.ai Founding Team(left-to-right): Peter Chen (CEO), Pieter Abbeel (President and Chief Scientist), Rocky Duan (CTO), Tianhao Zhang (Research Scientist)

 Embodied Intelligence, which was honed in on robotic grasping. Their press release explained it this way:

“We are building technology that enables existing robot hardware to handle a much wider range of tasks where existing solutions break down, for example, bin picking of complex shapes, kitting, assembly, depalletizing of irregular stacks, and manipulation of deformable objects such as wires, cables, fabrics, linens, fluid-bags, and food.”

Morphing into Covariant.ai in January of 2020—together with $40 million from investors—the four now cast themselves as creating the covariant brain: “a universal AI that allows robots to see, reason, and act on the world around them.” The Covariant.ai website further explains:

“Instead of learning to master specific tasks separately, Covariant robots learn general abilities such as robust 3D perception, physical affordances of objects, few-shot learning and real-time motion planning.”

And their AI must be mastering things pretty well. “ABB signed a partnership with Covariant in February, following a picking and sorting test held by ABB last year in which Covariant outperformed 20 other systems. In March, Covariant and KNAPP [the German logistics warehouse giant] signed a partnership to release a picking robot solution.”

A pre-AI robot would be lost in this roiling world of speed and intelligence.

Six thousand miles west of Dexterity and Covariant.ai, XYZ Robotics (Shanghai, global headquarters 2019; founded in Boston 2018) is pioneering the same AI territory with their motto: “Pick anything. Place anywhere.”

Mixing machine learning and robotic manipulation together, XYZ Robotics develops AI-enabled robotic perception and manipulation that doubles the speed of put-wall sorting and goods-to-person picking.

“The company hopes to use hand-eye coordination technology as the core, combining artificial intelligence, 3D perception and the advantages of robotic control technology, to provide visual picking robot products for logistics and industrial scenes.”

 

Rosen Diankov, the Bulgarian/American/CMU grad and expat, lit out for Tokyo and the JSK Robotics Lab at the University of Tokyo to build intelligent humanoids. In 2011, he co-founded Mujin.

The company’s strategy was to build robot controllers and camera systems and then integrate them with existing industrial robot arms.

The key products here are the controllers — each about the size of a briefcase, one for motion planning and one for vision — that act as an operating system that can control the hardware from any robot manufacturer. If a goal such as grasping an object is input, the controllers automatically can generate motions for robots, eliminating the traditional need to “teach” robots manually. The result, according to the company, is higher productivity for users.”

As Diankov says about the smart machines Mujin is building: “The approach is like that of a train, plane or rocket — you don’t want it to be self-learning, just predictable when it goes from A to B. That’s how you create innovation, with perfectly predictable systems. That’s what we’re trying to do with robotics. I like to call this machine intelligence, not artificial intelligence.”

Mujin’s gear is in operation at one of JD.Com’s warehouses (Shanghai), formerly run by 500 workers. Today, the same warehouse needs but five, even with the global acceleration in e-commerce.

Amused with all of this new-found innovation in a robot’s brain-hand-eye coordination is the godfather of intelligent pick ‘n place, Simon Melikian, the CEO of Recognition Robotics, founded in 2007 (Elyria, OH). He developed CortexRecognition software that is capable of guiding an industrial robot in a full 6 degrees of freedom (X, Y, Z, Rx, Ry, Rz).

“CortexRecognition is visual recognition and guidance software that works…very similarly to the human visual process and can be taught, and learn a large number of objects, then recognize and locate any of the learned objects in the camera’s field of view…Giving blind robots true sight and hand-eye coordination.”

As Melikian explains it: “The algorithms are based on the human cognitive ability to recognize objects. When I developed these algorithms, I was mimicking the human visual cortex. We’ve used knowledge about human brain function, the point of view of the human eye, and put this into the software realm.”

Robots marked post-AI
The intelligent warehouse and its AL/ML robot armies is the place where Mordor Intelligence sees global sales of warehouse robotics going from $6.12 billion in 2019 to $25.8 billion by 2025, which is a mega-jump for the next five years. The “emergence of connected systems,” says Mordor, “are helping industries perform various tasks, such as material batching, picking, ordering, packaging…and helping improve the operational efficiency by exponential margins.”

Like their human counterparts, pre-AI robots may well be facing an uncertain future of huge layoffs.

Close Menu
Subscribe