Close Ad

The Spoon

Daily news and analysis about the food tech revolution

Robotics, AI & Data

Gatik Gets $9 Million (CAD) to Winterize its Autonomous Middle-Mile Delivery Tech

Gatik, which makes autonomous delivery vehicles for the middle-mile, announced today that it has received $997,706 million CAD (~$788,511 USD) from Ontario’s Autonomous Vehicle Innovation Network (AVIN) R&D Partnership Fun, along with $8 million CAD (~$6.32 million USD) in unspecified “industry contribution.” The new funding will go towards winterizing Gatik’s autonomous driving technology.

Gatik develops self-driving delivery trucks for the middle mile, which typically means between two points within a company’s network, e.g. between a warehouse and a store.

The Ontario government will help Gatik’s autonomous driving technology withstand inclement weather. Right now, a lot of self-driving pilots and tests happen in sunny climates such as Arizona, Texas and California. Bright, sunny weather makes it easier for self-driving vehicles to navigate because road conditions are dry and the surrounding environment is clearer for the vehicle’s systems to “see.”

But if self-driving technology is ever to reach mass market scale, it must be able to operate in all kinds of weather. Not only will autonomous vehicles need to “see” in rain and fog and snow, they will also have to safely drive on wet and icy roads. Another self-driving delivery startup that has “ruggedized” their vehicles for harsh conditions include Refraction, which operates out of Ann Arbor, Michigan.

Gatik already operates a small fleet of autonomous delivery trucks in Canada. Last November, Canadian grocery chain Loblaw started using a five Gatik trucks to run food between automated picking facilities and retail stores. Gatik, which is headquarted in Palo Alto, CA, has been expanding its Canadian presence. The company recently moved into a 12,000 sq. ft. research facility in Toronto and expects to double its workforce there over the coming year. As part of its funding arrangement, AVIN will help Gatik attract and retain engineering talent in Ontario.

Last month, the company debuted its first electric delivery vehicle, which has a range of 120 miles and takes only 1.5 hours to charge. The company will also be running two delivery routes for Walmart in Arkansas and Louisina. The Arkansas route will go completely driverless (i.e., no human backup) this year, and the Louisiana route will be the first to use Gatik’s electric trucks.

Operating only within the middle mile makes it easier for Gatik to bring its autonomous driving tech to market. By focusing on the middle mile, the delivery trucks only need to navigate between two fixed points. By avoiding consumer delivery (the so-called “last mile”), Gatik limits the number of variables its trucks will encounter on a given route.

This narrow, middle-mile approach combined with winterized driving capability could give Gatik a huge boost in getting to more markets quickly.

McDonald’s May Sell Part of Tech Company Dynamic Yield

McDonald’s is considering selling part of AI startup Dynamic Yield, which it acquired in 2019, according to a report from the Wall Street Journal. 

The original intention behind the $300 million Dynamic Yield acquisition was to install the company’s AI tech into self-service kiosks and drive-thru menu boards at McDonald’s locations. With the implementation, menus would become more technologically sophisticated, able to offer things like more personalized recommendations for customers. McDonald’s currently has Dynamic Yield implementations in the U.S., Canada, and Australia.

But as WSJ reports, after analyzing the Dynamic Yield platform, McDonald’s has found that the tech “hasn’t delivered the promised sales boost” originally hoped for at the time of the deal. The QSR mega-chain aimed for Dynamic Yield order suggestions to boost drive-thru sales by 1 percent in the U.S. Sources told WSJ that sales have fallen short of that target, and one franchisee said that “the return on investment is just not there.”

The Dynamic Yield platform is just one piece of technology McDonald’s franchisees have called into question recently. Despite the chain’s aggressive push towards digital ordering and all the accompanying technological changes, franchisees have pushed back on how much they must pay for all this tech. Ernst & Young is currently conducting an audit of these technology fees, while franchisees have paused all non-essential communication with McDonald’s corporate.

Still, missed sales targets and franchisee tensions don’t spell the end of Dynamic Yield’s platform as part of Mickey D’s operations. McDonald’s said it would keep parts of the company, including those that service the chain’s drive-thrus and kiosks. While details are few and far between, McDonald’s shift towards stores with more drive-thru lanes and less dining room space suggest Dynamic Yield’s tech will continue to provide at least some value going forward. 

McDonald’s is currently “exploring” the sale of a portion of Dynamic Yield. There is no set timeframe for a deal, and it is entirely possible that no deal at all will happen.

Cecilia.ai Mixes Chatbot Capabilities with Its Robot Bartender

The robot bartenders we’ve covered so far at The Spoon are either just autonomous, articulating arms (Glacierfire), or high-volume vending machines (Rotender). Cecilia.ai, the latest entrant in the autonomous cocktail-making space, went live today and “she” is serving drinks a twist.

In addition to automated drink-pouring, Cecilia.ai features chatbot functionality to have a “conversation” with a customer. Cecilia looks like something out of a Vegas. The machine sports a big screen with a CGI female bartender. Walk up to Cecilia and start talking to order your drink. The Cecilia.ai website provides a sample conversation that goes something like this:

CUSTOMER: Hi there.

CECILIA: Hello there, what can I get you to drink?

CUSTOMER: Any recommendations?

CECILIA: My favorite is the Rusty Ale, but it’s a bit strong.

CUSTOMER: Do you have something sweeter?

CECILIA: Sure! Try the Green n’ Tonic.

And… scene.

According to website, Cecilia can make 120 drinks per hour (with reduced chit-chat), offers conversational script customization to fit a location, hundreds of available mixes and hold 70 liters (it doesn’t specify booze or mixer ratios). Voice control means ordering is contactless, and the large screen can be used for advertising purposes.

We’ve reached out to the company to find out more details like pricing and availability, and will update this when we hear back.

There are actually a number of robot bartenders coming to market right now. In addition to the aforementioned Glacierfire bar and Rotender robot, MSC Cruises is installing a robot bartender on one on of its ships, Macco’s robot is serving beer in Spain, and Makr Shakr continues to sell its robot bartender solution.

One reason for all this automated mixology is probably, like so many other things, the pandemic. Having hundreds of strangers yell out their orders into the faces of human bartenders isn’t such a great idea any more, thanks to COVID. A robot bartender eliminates that vector of human-to-human transmission. But another reason is that robot bartenders are machines that can crank out drinks around the clock without taking a break, and they do so without spilling or overpouring (which may suck for customers but is good for a bar’s bottom line).

From what we can gather, Cecilia isn’t aimed at high-volume nightclubs and bars, but instead is more for hotels, airports, VIP lounges, etc. This makes sense since a crowded bar blasting music in the background would make it difficult for Cecilia to hear a patron order a Patron with only their voice.

The only question that remains is whether sage wisdom and funny anecdotes are programmed into her chatbot capabilities.

Honeycomb.ai Helps Those with Allergies Eat at Restaurants Through AI

Vancouver, Canada-based Honeycomb.ai opened up its artificial intelligence (AI)-based web app this week to help diners with food allergies and specific diets navigate restaurant menus. The first version of the service is currently available for users in the U.S., Canada, and Australia.

From Honeycomb.ai’s website, a user can select from 29 different allergies or dietary preferences (only two filters can be picked at the same time). Then, after selecting a specific city, the website populates restaurants in the area and menu items that are acceptable based on the selected criteria. I spoke with Honeycomb co-founder Tamir Barzilai by phone this week to learn more about how the company’s algorithm functions. Tamir said the Honeycomb.ai team spent four years training the algorithm to learn what menu items are typically acceptable for different allergies and dietary preferences. Using the name, category, description, and any available metadata of a restaurant menu item found online, the algorithm is able to predict what ingredients will be found in it.

Tamir said that the algorithm is able to do about 90 percent of the work, but there is still some information that needs to be entered manually by the restaurant. After the algorithm makes its predictions, a restaurant must sign up on the Honeycomb platform to verify the accuracy of those menu predictions and add in additional ingredients that the algorithm may not be able to pick up on. For example, if a dish was cooked in butter, but the restaurant’s menu description does not directly state this, so a restaurant would add this to the database for a more complete breakdown of ingredients. Additionally, restaurants can add Honeycomb.ai’s plug-in feature directly to their website to help users find menu items suitable for their diets.

In the U.S., it is estimated that around 32 million people experience food allergies, while 36 percent of Americans follow some type of specific diet. In addition to Honeycomb.ai, there are various companies taking different approaches to make dining out with a food allergy or dietary preferences a more streamlined process. SevenRooms, a company focused on back-of-house restaurant tech, enables restaurants to track customer data to remember important things like food allergies and preferences. Nima, producers of handheld peanut and gluten detectors, conducted gluten and peanut tests on dishes at major restaurant chains and created a map of all “Nima-tested” restaurants.

Honeycomb.ai is currently available as a website and iOS app. The iOS app will be updated with a new version shortly, and the company will also be launching an app for Android in the next few months.

Future Acres Kicks Off $3M Equity Crowdfunding Campaign for its Ag Robotics Platform

Future Acres, a new startup developing a robotic platform for farms, announced today that it has launched an equity crowdfunding campaign that aims to raise $3 million.

The first product that Future Acres is building is Carry, a self-driving robot meant to, as the name suggests, carry crops around a farm. Carry uses GPS, computer vision and machine learning to autonomously navigate a farm (it can also be tele-operated), and can lug 500 lbs. of crops across all types of terrain and inclement weather. The electric robot has a 7 -- 10 hour battery life and can travel 6 -- 10 miles on a full charge.

Introducing Future Acres

If this sounds familiar, that’s because Carry is similar to Augean Robotics’ Burro, which also autonomously hauls up to 500 lbs of food and gear around the farm.

Future Acres is also looking beyond the simple act of carrying things and towards developing a true autonomous platform that can be used to perform other tasks around the farm like precision spraying, disease detection and, eventually, crop picking.

Farms in the U.S. face labor shortages caused by factors such as COVID restricting the movement of migrant labor and a patchwork of differing state and federal labor laws. Even if a farm is able to secure all of the workers it needs, that work is still hard and done under harsh conditions like extreme heat.

This is where farm automation can help. By automating some of the less skilled work involved with harvesting, such as carrying bushels around, human workers can focus on more delicate tasks such as picking, or coordinating logistical processes.

Right now, Future Acres has one prototype currently being tested. With the new funding it raises, the company will focus on developing the next version of Carry. Future Acres CEO, Suma Reddy, told me by phone last week that the company will work with farms to figure out what business model(s) work best, but right now, the Carry system costs between $800 and $1,200 a month for the hardware and software.

PIX Moving Raises Pre-Series A Round of Funding for Self-Driving “Skateboard” Chassis

PIX Moving, a Chinese company that makes chassis for self-driving vehicles, announced at the end of last week that it has raised a Pre-Series A round of funding. Details in the English-language press release provided to The Spoon didn’t provide a specific dollar amount, only saying it had raised “tens of millions” in Chinese Yuan. The only investor listed in this round is Guizhou Transportation Planning Survey&design Academe Co., ltd.

Dubbed the “skateboard,” PIX’s chassis can be used to power a number of different types of low-speed autonomous vehicles. Most relevant to Spoon readers, restaurant or third-party delivery companies could use affix lockers on top of PIX’s skateboard to make a mobile automat, or a grocer could create a temperature-controlled store on wheels.

According to today’s press announcement, 23 types of vehicles have been developed using PIX’s platform, and vehicles have been delivered to roughly 100 customers around the world. To help speed up production of its skateboard, PIX has also developed Lightsaber, a 3D metal printing manufacturing system that removes the need for molds.

When we last checked in with PIX, the company was still navigating the various regulations around self-driving vehicles in China and the U.S., though the company had deployed vehicles to private campuses in both countries.

While self-driving vehicles still have a number of hurdles to overcome before they go mainstream, they have gotten much closer to hitting the road, as it were, over the past year. Last year, Nuro’s self-driving pod-like vehicle got approvals from the federal and California governments to operate on public roads. More recently, Gatik, which builds autonous delivery trucks for the middle mile, announced that it would operate two of its routes without human drivers this year.

As it moves closer to market, PIX said it will use its new funding to mass produce its vehicles for a variety of applications.

Mars Wrigley Launches Mobile Robotic Kiosk at ShopRite

The impulse aisle in the checkout line has long been the bane of many parents’ grocery shopping trips. While cashiers scan items and parents wait to pay, racks of candy and treats are within arm’s reach of bored kids sitting in carts.

But if you thought tempting kids with treats out checkout was rough, wait until there’s a shiny new robot wandering the grocery aisles, offering up candy.

Mars Wrigley and Wakefern Food Corp. announced last week that they are working with robot company Savioke to deploy a mobile robotic vending kiosk at ShopRite store in Monroe, NY.

According to Kiosk Marketplace, the robot, dubbed “Smiley,” plays music and dances (doing the robot, we assume) (sorry!), while offering up treats and such for sale.

There aren’t a ton of details, such as how many treats Smiley can hold, or the mechanics of how the treats are dispensed and paid for. (We reached out to Mars Wrigley for more information.)

We’ve seen these types of robots before. Self-driving robots can wheel around inside existing retail spaces to act as promotional, err, vehicles, or direct avenues. In China, FANBOT is already scurrying around cinemas, malls and hotels, selling drinks, snacks and more. And Pudutech’s robot, which is basically a series of shelves on wheels, cruise around grocery stores in Japan and the Netherlands, showcasing items that are on sale.

At the risk of tooting my own horn too loudly, automated mobile kiosks was a trend I said to look out for in 2021. So I think we’re going to see a lot more robots roaming store aisles trying to sell us stuff. Maybe their LED faces will need to show a scowl though, to keep the kids away.

MyFitnessPal Adds AI Scanning Technology to iOS for Tracking Calories

MyFitnessPal this week announced the launch of its new AI-powered scanning feature that automates some of the process of identifying and tracking the ingredients in a user’s food items. The company has partnered with Passio, which provides computer vision for companies through its AI platform, to develop this new feature.

If you’ve ever tried using a food tracking app or website, you’ll know what a pain it is to enter every single ingredient of each meal and snack you consume. With MyFitnessPal’s scanning feature, a user selects the Meal Scan feature in the app, then holds their phone’s camera over the food. The AI scanning tool is able to identify the food, both the type and the amount of it, through the combination of Passio’s food recognition technology and MyFitnessPal’s massive database of 14 million foods.

This feature can help users avoid having to search for each ingredient and food within the database, but it is not yet fully automated. After scanning the food, the user must confirm the foods and amounts the app has identified. After confirming, calories, fat, protein in the food is calculated and this information will be automatically added to the user’s food diary.

That being said, using computer vision for tracking food can still save a user a lot of time, and MyFitnessPal is not the only company using something like this. Bite.ai is another company that has a free food tracking app, and it uses a similar technology that applies computer vision to identify food that is then added to the user’s food diary.

MyFitnessPal’s scanning technology is currently available on iOS, and it will be made available for Android phones in the near future. The MyFitnessPal app is free, but only those who are premium members for $9.99/month will have access to the scanning feature.

Imperium Drive Comes Out of Stealth with Bandwidth-Aware Teleoperation for Robots

One of the many questions facing delivery robot startups as they come to market is how much autonomy to give the robots. Should they go full autonomous driving, even though that is more technically complex and there is still a patchwork of regulation that needs to be dealt with? Or should they go with less autonomy and use humans to help guide or even drive their robots to sidestep some of the complications that come with self-driving vehicles.

For robot delivery startups wrestling with these questions, Imperium Drive says it’s here to help. Based in Europe (the company is scattered across different countries because of the pandemic) and part of the TechStars Smart Mobility cohort, Imperium Drive makes a teleoperation system for autonomous vehicles, including the small rover robots favored by the likes of Starship, Kiwibot, Postmates and others.

Imperium does the whole teleoperation stack, from the software onboard the robot to the human on the other end who helps the robot navigate. Imperium lets robot companies choose how much teleoperation they want, depending on their robot’s level of autonomy. Imperium can have a human simply monitor the robot remotely in case it gets stuck or runs into trouble, or the company can plot points on a map for the robot to autonomously follow. At the far end of the spectrum, Imperium can have a human actually drive the delivery robot remotely, like a videogame.

Imperium Drive Co-Founder and CEO, Koosha Kaveh, told me by phone this week that his company’s secret sauce is its ability to operate even when network connections provide only low bandwidth. As the robot runs around town, the strength of its cellular connection will vary, sometimes offering very small pipes for data to get through.

“We’ve developed our own AI predictive engine that predicts changes of network parameters,” said Kaveh, “And we change automatically our streaming engine based on availability.”

A easy way to think about Imperium’s bandwidth technology is Netflix. The movie streaming service will detect how much bandwidth you have (e.g., a cellular connection versus wired Ethernet) and serve a movie in a resolution fit for that situation. Imperium does the same thing, just with data from the robot.

Delivery robots are actually streaming a lot of data back to their headquarters. There’s video from the robot’s on-board cameras as well as lidar and radar information. Imperium adapts what is streamed based on the amount of bandwidth. For example, if there’s very little bandwidth, Imperium can send just wireframes of the robots surroundings. Kaveh said that it can stream the relevant information a teleoperator needs at under 1MB of data.

The idea of teleoperating robots brings up the question of scale. It’s easy to understand self-driving robots scaling up to meet demand because that’s the whole point — there is no human labor to pay. Once you have the robots, they can just run around the clock with no additional cost. What happens to the economics when you have a human handling a robot?

Kaveh says Imperium has a network of inexpensive labor in Eastern Europe that it can tap into to teleoperate robots. And Imperium isn’t alone in using humans to guide robots. Kiwibot has a team of people in Colombia that plots the courses for its robots (not full-on driving). And Tortoise skipped the idea of self-driving altogether for its robots, believing it can create a Mechanical Turk style army of human gig-work teleoperators.

The fact that Imperium Drive exists is at least some indication that the delivery robot space is maturing. As we outlined in our Delivery Robot Market Report, there are many companies around the world deploying delivery robots on city streets. Imperium Drive is part of a typical business cycle for new market categories like delivery bots. It’s not creating the robots themselves, but adding a layer to make those robots run more efficiently. As robots gain traction, we’ll see more third-party add-ons like this meant to improve robot delivery operations.

Rotender is a Drink Making Robot Built for High-Volume Bars

I worked at a nightclub in college and one thing I remember from that experience was the sheer volume of drinks bartenders poured each night. Our job wasn’t about fancy bottle flipping a la Cocktail, just getting drinks to the consumers and their money in the till.

This high-volume approach is what’s driving the team at Rotender, which has built a robotic vending machine that serves drinks. Each Rotender holds 16 one-liter bottles and five different types of syrups (e.g., cranberry or orange juice), has an automated soda gun and makes it own ice. Once up and running, the Rotender can serve a drink in 15 seconds and make more than 350 drinks before needing to be refilled.

Rather than sitting behind the bar like SomaBar, the Rotender is actually meant to be installed where consumers can use it. A customer uses their mobile phone to scan a QR code on the Rotender, which brings up the drink menu in the Rotender app. Users select their drink, pay for it, place their glass in the machine and then scan a QR code again to ensure that they are by the machine to pick up their drink. Once that’s done, the Rotender mixes the drink and serves it.

Right now, Rotender relies on the venue to handle age verification, though the company is exploring existing software solutions to handle that. The machine also keeps track of how many drinks it serves each customer. Should a consumer order too many drinks in a single hour, Rotender can pause service to that individual so as not to over-serve them.

When we talk about robotics, there is typically a discussion around how robot-like any machine should look. Other robo-bartenders on the market serving drinks at Glacierfire in Iceland or The Tipsy Robot bar in Las Vegas use articulating arms to make the drinks. Rotender has instead forsaken this “robot” approach in favor of creating more of a straight up drink-making machine. As Rotender Co-Founder and CEO, Ben Winston explained to me by phone this week, the reason is speed.

Rotender is built to churn out drinks for high-volume locations. As such, the machine focuses on meat and potato drinks — rum and coke, gin and tonic, etc. More complicated drinks are left to the human staff at the bar.

By foresaking expensive articulating robotic arms, Winston also said that Rotender is a more economical option for bars. It costs $999 a month to lease a Rotender (a robotic arm on its own can cost tens of thousands of dollars). In addition to that lease, Rotender also adds a per drink charge that it determines with the venue.

Rotender is about to do a pilot with a bar in Los Angeles, and Winston said they can currently serve clients in California and Nevada. Sadly, though, the Rotender won’t be doing any Cocktail-style bottle flips.

Standard Cognition Raises $150M Series C for its Cashierless Checkout

Cashierless checkout startup Standard Cognition announced today that it has raised a $150 million Series C round of funding. The round was led by SoftBank Vision Fund 2, with participation from existing investors CRV, EQT Ventures and TI Platform Management, along with new investors SK Networks and others. This brings the total amount of funding raised by Standard to $236 million.

Standard Cognition creates cashierless checkout experiences for retailers using a system of cameras, computer vision and artificial intelligence (AI). Once installed and integrated with a payment system, customers can walk in, grab what they want, and leave without needing to scan items or wait in a checkout line.

Shopping in seconds with Standard, the world's most advanced autonomous checkout solution

This round of funding is far and away the largest we’ve seen for a cashierless checkout startup. Similar startups that retrofit stores with cameras and AI include Grabango ($32 million), Trigo ($89 million), and Zippin ($12 million). Also worth mentioning is Accel Robotics, which raised $30 million from SoftBank the company, not SoftBank Vision Fund.

In its press announcement, Standard said this new funding will help the company scale up its solution. Standard said it has already been working with customers including Alimentation Couche-Tard, Inc. (parent company of global convenience store brand “Circle K”) and Compass Group. Standard said it will outfit “hundreds” of checkout-free stores this year and has a goal of more than 50,000 stores over the next five years.

It is an understatement to say that it’s been a busy year so far for cashierless checkout. Just today, cashierless startup AiFi announced a partnership with Dutch convenience store chain Wundermart that includes plans to open up 1,000 checkout-free stores. Previously, New Zealand-based IMAGR partnered with Japanese retailer H2O, Berlin-based Nomitri came out of stealth, and Zippin launched a checkout-free store in a hotel in Japan.

A big reason for all this activity (and investment) is the ongoing global pandemic. Retailers are looking for ways to alter the shopping experience in a COVID world, which includes reducing the amount of human-to-human interaction the happen at the store. Removing the cashier removes one vector of transmission, and removing the checkout line altogether means that fewer people congregate inside a store.

In other words, expect to see more cashierless checkout announcements (and investments) in the coming months.

AiFi and Wundermart Partner up for Cashierless Checkout Stores

AiFi is bringing its computer-vision based cashierless checkout technology to Amsterdam-based Wundermart‘s chain of convenience stores worldwide, the two companies announced today.

Wundermart currently operates 70 unattended “grab-and-go” convenience stores throughout Netherlands, Belgium and Germany. These stores are typically between 50 and 200 sq. ft. and located in hotels, offices and transportation hubs. While the stores themselves are unattended by staff, Wundermart shoppers currently need to manually scan each item and make their payment at a self-checkout kiosk.

Through this new partnership, Wundermart will incorporate AiFi’s camera-based checkout system. As shoppers enter a Wundermart store, they will need to scan the store app or swipe their credit card at a kiosk. Once inside, AiFi’s camera+computer vision system will automatically keep track of what people take and then charge them automatically as they leave the store — no product scanning needed.

In addition to its computer vision checkout, Wundermart will be implementing AiFi’s OASIS system, which uses computer vision and sensors to track products in the store for more automated inventory management.

Wundermart will also be integrating AiFi’s technology into its Wunderware SaaS package. Wunderware is sold to third-party retailers so they can create their own autonomous stores. According to today’s press announcement, Wundermart and Wunderware plan to open more than 1,000 locations equipped with AiFi’s technology.

The COVID pandemic is pushing retailers of all sizes to re-evaluate the number of touchpoints they have in their stores. Cashierless checkout provides a more frictionless shopping experience that reduces the amount of human-to-human interaction. Not only does it remove the cashier as a vector of viral transmission, but early data suggests that shoppers using cashierless systems spend less time in a store, so there are fewer people congregating in checkout lines and aisles.

In addition to more frictionless checkout, Wundermart said it will be using AiFi’s system to also learn how people are shopping inside their stores. While the company’s system doesn’t use facial or biometric data to track someone, it can analyze what items people pick up (and put back) to help a store operator determine how to most efficiently set up products.

Only a couple of months in and 2021 has already been a busy year for cashierless checkout. New Zealand-based IMAGR has launched its smart-cart solution with retailer H20 in Japan. German startup, Nomitri came out of stealth with its smartphone-centric, cashierless checkout. And Zippin launched an unattended convenience store at a hotel in Japan.

For its part, AiFi has said previously that it plans to have deployed 330 new and retrofitted autonomous stores by the end of 2021. With the Wundermart partnership, they are certainly on their way to achieving that goal.