Home Robotics Brightpick CEO Jan Zizka discusses growth of robot-to-goods system

Brightpick CEO Jan Zizka discusses growth of robot-to-goods system

Brightpick CEO Jan Zizka discusses growth of robot-to-goods system


Take heed to this text

Voiced by Amazon Polly
image of a lego model of the brightpick mobile robot.

Brightpick applies machine imaginative and prescient and AI to robotic choosing. Credit score: Mike Oitzman

The Robotic Report Podcast not too long ago featured a dialog with Photoneo’s Jan Zizka in regards to the new BrightPick achievement resolution that the corporate launched earlier this yr. Under is an edited transcript of the dialog.

Headshot of Jan Zizka

Jan Zizka, co-founder and CEO of Brightpick. | Credit score: Brightpick

Zizka is the co-founder and CEO of Brightpick and its guardian firm, Photoneo Brightpick Group. He has filed greater than 20 patents, starting from 3D sensing strategies to cell robotics. Zizka can also be a acknowledged skilled within the fields of synthetic intelligence, machine imaginative and prescient, and warehouse automation.

Photoneo’s clients embrace Normal Motors, Volkswagen, and KUKA. Earlier than co-founding that firm, Zizka was a analysis engineer at Micro-Epsilon, a number one sensor producer. 

He earned a Ph.D. in pc imaginative and prescient from Comenius College, throughout which he additionally did analysis on computational cameras and optics on the Massachusetts Institute of Know-how.

What’s Brightpick? What forms of market functions or particular warehouse workflows is it designed for?

Brightpick is a spinoff from Photoneo. The answer leverages the 3D notion and robotic choosing expertise that’s the core of Photoneo. It’s a warehouse automation resolution particularly designed for order achievement and focusing on completely different verticals like e-grocery, third-party logistics, and e-commerce.

The Brightpick resolution is optimized to deal with smaller objects like small electronics, grocery gadgets, and related gadgets that may be saved in bulk in stock totes. 

Describe the standard achievement workflow for a Brightpick autonomous cell robotic (AMR).

The Brightpick AMR travels by the warehouse because it builds one discrete order much like a human picker. The warehouse stock must be organized right into a set of totes and the person gadgets must be “toteable.”

Leveraging an onboard cell manipulator, the eaches choosing is completed immediately from the majority stock totes pulled from the shelf whereas the BrightPick AMR is stationary within the warehouse aisle.

What typical measurement of totes can Brightpick deal with?

We prefer to have one uniform-size tote for the warehouse. The popular totes are often 60 by 40 cm [23.6 by 15.7 in.], and the peak of the tote may range, relying on the gadgets saved. The totes could be cut up into a number of compartments. The storage necessities depend upon the order frequency of various gadgets. Area utilization is essential, and having compartmentalized totes helps to enhance the effectivity of the answer.

wide angle view a warehouse with multiple Brightpick Picker AMRs running around pulling order items. | Credit: Brightpick

Brightpick picker AMRs traverse the warehouse to tug particular person gadgets for buyer orders. | Credit score: Brightpick

What was the design course of and the “Aha” second for the Brightpick resolution? Photoneo’s energy is in 3D imaginative and prescient, 3D notion, and robotic imaginative and prescient steering, particularly for functions like bin choosing.

Through the years, we have now been learning how basically troublesome achievement is, and we determined it’s a large enough downside to be solved. We made this robotic from the bottom up using all of the Photoneo applied sciences into Brightpick.

We already had all the vital capabilities to perform this troublesome activity. Beginning with our 3D cameras, over time, we developed our machine studying engine.

In grocery, we will decide round 80% of things that we see from the tote. We prefer to name it “totally automated achievement,” which isn’t fully true, however our definition of totally automated achievement is in precept that we will automate 95% of your entire course of. 

What are among the points that you simply’ve needed to clear up for this new automated workflow for achievement?

Orchestrating the robotic is a troublesome downside. We use particular algorithms within the classical “touring salesman downside” to ship the person robots out for every SKU-picking activity. If it could be only one robotic, it could be easy.

One of the complicated issues we’re fixing is one thing we name time-space planning. When you could have a fleet of robots, every robotic may simply clear up a touring salesman downside individually, however right here we have now to orchestrate the robots as a fleet.

One difficulty we have now is that the robots is perhaps competing for a given aisle, or perhaps a given SKU, and we have now to keep away from that. So it is a tremendous complicated downside. Because the robots journey by the warehouse, we have now to plan their paths, plan the choosing dwell occasions, and negotiate conflicts the place two robots need to be in the identical place on the similar time.

SITE AD for the 2024 RBR50 call for nominations.Submit your nominations for innovation awards within the 2024 RBR50 awards.

There are another robots available on the market, like InVia’s AMRs, which pull stock totes from a shelf and transport them to an item-picking location, then return the stock to the warehouse shelf. Is that this much like an automatic storage and retrieval system (ASRS)? Do you categorize Brightpick as an ASRS-type system?

I would categorize the InVia resolution as a goods-to-person [G2P]-type resolution, like Kiva Techniques. The opposite rising class is “robots to items,” and perhaps that’s the appropriate class for Brightpick.

With the Kiva-like options, the robots have to maneuver bulk gadgets lengthy distances simply to choose a single SKU for a buyer order. However an important half is that the robotic has to carry this tote again to its long-term storage location within the warehouse. So bulk gadgets are constantly moved to the picker, again to the stock location, and again to the picker, and so forth. 

So we imagine that the BrightPick resolution basically and mathematically requires 40% to 50% or much less journey. And that signifies that you might have fewer robots deployed on the ground. Our aim is to create a super-efficient system. We now have to attain these efficiencies.

Autopicker pulls bulk item totes from the warehouse shelf to singulate individual customer items

The Brightpick Autopicker AMR pulls bulk merchandise totes from the warehouse shelf to singulate particular person buyer gadgets. | Credit score: Brightpick

Brightpick makes use of a SCARA manipulator to choose and place particular person gadgets from totes. Why a SCARA robotic versus a six-axis manipulator?

Possibly this isn’t the fascinating reply, but it surely comes right down to value. We made a trade-off between the complexity of AI and algorithms and the price of the manipulator. The selection to make use of a SCARA manipulator for selecting signifies that the AI, gripper, and algorithms behind it are extra complicated for the bin-picking operation.

It took us just a few years, and we designed round 200 grippers with completely different variations and completely different methods. What we ended up with is one thing that may be a good stability between complexity, value, and capabilities.

Immediately, we can’t be 100% assured of choosing each merchandise, however then no resolution can. I believe people can nonetheless function a fallback resolution by participating a close-by affiliate to assist decide an merchandise. I might say for purchasers, end-to-end reliability is rather more vital than the truth that you’ll be able to decide like 99.9% of the gadgets.

Do you utilize a single gripper for all of the gadgets, or are you doing a software change for various sorts of components?

We now have we have now each ideas. We do quite a lot of completely different gripper ideas together with a number of suction cups, gripper exchangers, and suction-cup exchangers.

So all of that is doable. It seems like not each warehouse requires such flexibility. Finally, we are going to find yourself with one thing like a gripper exchanger with two or three fingers.

So that you outlined this new phase a couple of minutes in the past as a “robot-to-goods” resolution. What makes this kind of system viable now? 

The robotic you see right here might be the fifth technology. It’s not designed to do 50 various things; it’s not like a general-purpose humanoid. Alternatively, what we have now carried out is optimize the price of every aspect of the robotic to scale back the price of the required performance.

I’m tremendous pleased with what our workforce was capable of do on the {hardware} design facet and with procurement. 

screenshot of the Brightpick fleet management software.

Brightpick contains software program and optimization algorithms to manage materials stream and the operation of all the particular person robots on the warehouse ground. | Credit score: Brightpick

Bin choosing with cell manipulation has been a “holy grail” downside. Is there a breakthrough on this design that you simply’re notably pleased with?

It comes right down to a holistic view of all of those newest developments in synthetic intelligence. It’s a must to have the appropriate gripper, the appropriate path planning, the appropriate AI, and the appropriate idea. Every little thing needs to be excellent. And if one half is lacking, abruptly, it’s not possible. 

On the AI facet, this was not doable 5 to 6 years in the past. The intense breakthroughs allow machine studying to generalize to unseen gadgets. I believe that that is so highly effective. That is just the start.

For instance, IAM Robotics struggled for years to make cell manipulation viable. Why do you suppose its method wasn’t profitable? 

I believe it’s most likely two issues. First, should you purchase a six-axis robotic from a significant provider, you will get an excellent product. However the worth is prohibitive. I believe there’s merely no approach which you can make it occur as we speak. 

Second, IAM Robotics was choosing immediately from cabinets. That is complicated a really expensive replenishment course of. Our replenishment resolution is to simply dump it into the tote. 

Replenishment in our system might be like 5 to 6 occasions extra environment friendly as a result of robots will slot a given tote into the appropriate place within the warehouse. We try to research the frequency at which gadgets are picked to arrange the place gadgets ought to go on the shelf.

You’ve constructed an algorithmic achievement heatmap of frequent gadgets. Are you continually reorganizing the warehouse to make it extra environment friendly for the robots to tug gadgets?

Sure, precisely. In classical G2P programs, you should decrease the gap you should journey with the robots. So shifting high-frequency SKUs nearer to the human pickers could be very apparent right here.

With these older G2P robots, they’re all equal, standalone, and self-sufficient. All of the robots can do all of the processes. They’ve their trajectories.

It’s not so apparent now to ask: “Do I need to create a hotspot someplace?” Most likely not, as a result of I can steer too many robots to the identical aisle or similar spot, and this congestion creates a brand new downside.

It’s a must to someway stability in an enormous warehouse, between shortening distances and spreading these gadgets out too far. It’s a key optimization downside.

image of a brightpick autopicker in a warehouse aisle.

The Brightpick Autopicker is designed for slim aisles. | Credit score: Brightpick

Do you allow area within the aisles for two-way visitors to keep away from potential collisions or path conflicts?

Two years in the past, a lot of the aisles have been one-way. Immediately, nearly all of them are two-way. With robust planning algorithms, we will determine the paths for the robots to keep away from potential conflicts within the aisles.

We strive solely to have one robotic in every aisle at a time. However this time-space planning downside is a tough downside to unravel. There are a number of ranges to it. However security with robots and collaborative manipulators is important. 

Are you able to share a few of your early buyer deployments?

Our largest buyer as we speak is Rohlik Group, one of many fastest-growing e-grocers in Europe. These are large orders and largely ambient gadgets. We’re deploying a calming phase within the upcoming weeks.

Grocery could be very fascinating. In case you think about your loved ones shopping for 100 gadgets per week, that’s quite a lot of picks per order. In case you evaluate that to different e-commerce, it could possibly be that groceries are greater than your entire e-commerce phase. I believe groceries are one among our prime targets. 

We’re ending a brand new deployment for a pharma consumer. On this use case, it’s quite a lot of smaller packing containers and a excessive variety of SKUs. Once we have been designing the Brightpick system, this was one thing that we had in thoughts.

When an software, we take a look at the variety of picks per line. For instance, should you ask me for an software having 100 gadgets per line, meaning quite a lot of picks for the given SKU.

Sure, that’s most likely not optimum, and no human is far sooner. We’ve been counting 100 gadgets, you’ll nonetheless do it 5 occasions faster than the robotic. We’re all the time looking for one of the best match so as profile.

Take heed to the interview in its entirety right here:



Please enter your comment!
Please enter your name here