How iRobot used data science, cloud, and DevOps to design its next-gen smart home robots

iRobot has used its new design, software, and data science strategies to expand into new areas, using an approach to the smart home that is different from its big tech rivals.

iRobot 2.0: How the company set itself up for its next-gen robots iRobot launched the Roomba S9 and Braava Jet M6 with plans for a lawnmower on deck. It took a lot of IoT, data science, product design and cloud to make it all happen. Read more: https://zd.net/2K2L8iU

Watching iRobot's new vacuum tag-team with the company's latest robot mop to clean a room is to glimpse a Jetsons-like orchestration of the home of the future.

s9-m6-photo-insitu-imprintlink2.jpg

iRobot's Roomba S9 and Clean Base with the Braava Jet M6. 

iRobot

Seeing the two working together shows how smart home robots are starting to talk to each other, in this case through a technology called Imprint Link, and coordinate to be more effective without needing human involvement. But these two new bots, and the ones that will follow them, also represent the evolution of iRobot's approach to software, data science, and design.

Certainly, the Roomba S9 (the vacuum) and Braava Jet M6 (the mop) represent a breakthrough. Using Qualcomm processors, smartphone-level processing, more memory, and Imprint Link, the two robots work together. When Roomba gets back from a vacuum job, it lets the Braava know when to start. The two robots share maps of the home, and you can command them to tackle jobs and prioritize via iRobot's app.

irobot-pdf-cover.jpg

Download this article as a PDF (free registration required).

Here's where you'd expect me to hit you with specifications and features like Roomba's new rubber brush with fletches to pick up larger debris, advanced 3D sensors, and for the S9 price of $999 and S9+ price of $1,299, the Clean Base Automatic Dirt Disposal -- a system that empties the robot and puts dust and debris into a bin. I might as well also mention the Braava Jet M6 price of $499, and that you can buy the Clean Base separately for $349.

But the far more interesting tale revolves around how iRobot, which is almost 30 years old and has sold more than 25 million robots, found a new focus and leveraged data science to become a significant player helping to decide how the smart home will operate. If you're going to promise consumers that they won't have to touch their Roombas or worry about vacuuming for months at a time, there's a lot of backend IT and architecture work, data science, and cloud computing behind the scenes. This data-driven and DevOps approach may just set up iRobot as a key smart home leader that can ride emerging trends such as aging in place.

Download this article as a PDF (free registration required).

iRobot 2.0

colin-angle-ecosystem-of-robots.jpg

iRobot CEO Colin Angle with his ecosystem of home robots. 

iRobot/Webb Chappell

For iRobot CEO Colin Angle, the launch of these new robots is the first product cycle that represents iRobot 2.0. That includes a new design language, a common software platform and the plan that the company will move beyond its core home vacuum and mopping systems: iRobot is launching the Terra robot lawn mower in beta in the US and commercially in Germany.

"These robots are the beginning of a new business phase for iRobot. From the design language, to spatial understanding as a foundation, what we are doing at the company is moving from 'hey buy this robot' to 'hey enjoy this service,'" says Angle.

The core item in the new design language is the circle in the middle of the robots. The circle represents the history of iRobot, which featured a bevy of round Roomba robots. "The circle is a nod back to the round robots and gives us the ability to be more expansive with geometries," he explains.

But iRobot 2.0 also represents the maturation of iRobot. "Innovation at iRobot started back in the early days with a toolkit of robot technology. Innovation was really about market exploration and finding different ways for the toolkit to create value," Angle says.

Through that lens, iRobot explored everything from robots for space exploration to toys to industrial cleaning and medical uses. "Our first 10 to 15 years of history is fraught with market exploration," Angle says.

Ultimately, iRobot, founded in 1990, narrowed its focus to defense, commercial and consumer markets before focusing solely on home robots. iRobot divested its commercial and its military robot division, which was ultimately acquired by FLIR for $385 million

Here's a brief history:

- 1998: iRobot developed military robots under a DARPA research contract.
- 2001: iRobot's PackBot military robot searches at World Trade Center after Sept. 11 terrorist attacks.
- 2002: Roomba launched.
- 2005: iRobot went public.
- 2012: iRobot acquired Evolution Robotics, which made automated floor mopping robots.
- 2016: iRobot sold its military robot business to Arlington Capital Partners and the unit becomes Endeavor Robotic Holdings, which was acquired by FLIR Systems in 2019 for $385 million.
- 2016: iRobot launches Braava jet mopping robot.
- 2018: iRobot launches Roomba i7+ with Clean Base Automatic Dirt Disposal and the ability to store maps of the home.

The next phase for iRobot revolves around new markets starting now. "We had to decide whether iRobot was a Roomba company or a consumer robot company. We decided that we would extend well beyond Roomba and invest in technology and platforms that would allow us to move into logical adjacent markets," Angle says.

The latest Braava Jet represents the logical extension of iRobot's core expertise as does Terra, iRobot's lawn mower that will launch in the second half of the year in Germany. The US will get a beta program for Terra in the second half of 2019.

Angle says customers have been very forthcoming about what markets they've wanted iRobot to enter. "The consumer demand was deafening," Angle says. "The big question is how to do this safely."

Download this article as a PDF (free registration required).

Wall Street analysts are mostly positive on the prospects for Terra and iRobot's potential as a smart home orchestrator. Jed Dorsheimer, an analyst at Canaccord Genuity, said in a research note that Terra "represents another Roomba-like opportunity for iRobot," but notes it will take at least a year for Terra to gain traction.

Currently, 90% of iRobot's revenue is Roomba-related with Braava expected to represent 10%.

How Terra will work

iRobot started from the base of core demand and decided to leverage its mapping know-how. Many of iRobot's indoor robots use technology called iAdapt to store maps of a home. If, the company figured, it could do something similar outdoors with Terra, it could launch a safe lawn mower.

"There were a parade of constraints and we spent a couple of years evaluating technologies for an outdoor navigation system," Angle says.

irobot-terra-hero.jpg

iRobot

Also: Robotics in business: Everything humans need to know

Most robot lawn mowers require an electric fence around the yard to keep the machine contained. Angle said that approach was complicated. Ultimately, iRobot decided ultra wideband wireless on the 6.5 GHz spectrum would work: the robot uses outdoor beacons to know where it is.

Why can't Terra use the camera systems deployed by Roomba? Angle explains that a camera or visual system on a robot would suffer from saturation from direct sunlight and changes in the visual environment. "The consequences of losing localization are much higher with a robot," Angle says.

irobot-terra-navigation.jpg

iRobot


Terra and Roomba are different in how they use landmarks. A Roomba can use visual landmarks, but Terra relies on beacons that are placed into the ground.

However, the two robots share code for the navigation stack. "There are opportunities for reuse in the code," Angle says.

If iRobot is able to scale its product lines and leverage adjacent markets, growth should accelerate. iRobot's annual sales in 2016 were $660.6 million growing to $883.9 million in 2017 and $1.09 billion in 2018. Wall Street analysts expect iRobot to deliver revenue of $1.29 billion in 2019 and $1.48 billion in 2020.

Download this article as a PDF (free registration required).

Retiring tech debt

Those opportunities to reuse code and have a shared software stack could be the real key to the success of iRobot 2.0. Angle says the company had to unify navigation code as well as the operating systems between products. Previously, each product had its own unique spin on operating systems, platforms, and architecture.

"For a CEO the hardest thing to do is to allocate resources to pay down the technical debt column and make investments to move to a platform that has reuse and multi-robot architectures," Angle says. "We had to do it since it's fundamental to iRobot's strategic plan that we have a unified navigation experience. Once you learn iRobot, you can use all of our robots with full competence."

Getting to a common platform and architecture has taken the last five years, Angle explains. With each new robot carrying more software code and processing, iRobot can now send updates to products in the field.

Mike Tirozzi, chief information officer at iRobot and senior vice president of chief architect data solutions, says his position evolved from IT and enterprise systems to driving data-based products with analytics. To pull off that transition, Tirozzi needed an architecture that could leverage data and pivot as business needs evolved.

Tirozzi said that the cloud has allowed iRobot to retire tech debt, but the work is never done. "You never get rid of tech debt completely," he says.

Download this article as a PDF (free registration required).

Cloud computing is playing an increasing role behind the scenes. iRobot said in regulatory filings that it moved its infrastructure for its connected robots to Amazon Web Services. Indeed, Angle is a keynote speaker at the AWS re:MARS conference on artificial intelligence and robotics, and speaking on iRobot's first quarter earnings conference call, Angle says, "We will be collaborating with Amazon to demonstrate a full-scale smart home featuring Alexa-enabled iRobot devices."

Tirozzi says iRobot's architecture initially revolved around putting standard processes around enterprise applications and systems. "Then we started filling in holes like IoT, which was then called M2M (machine to machine). IT was running DevOps, but there was an opportunity to use data from the robots to help with product and customer service so we started building out a data platform," says Tirozzi. "We started a journey with data where we didn't know what we were going to initially use it for."

That journey led to a microservices and serverless architecture powered by AWS, a top cloud provider. Tirozzi says iRobot uses a bevy of services but serverless Lambda is a primary one. "Our microservices are based on APIs connected to data via four micro platforms,"Tirozzi says.

In September 2015, iRobot struggled to handle traffic to its iRobot Home App, which was handling data for its first connected Roomba. AWS was used for traffic bursting. iRobot began moving some of its workloads to the cloud in 2013 through another vendor, but had trouble scaling. iRobot uses more than 25 AWS services including AWS Lambda and AWS IoT Platform.

Here's a look at how Roomba connected to AWS architecture via a 2016 case study:

aws-roomba-architecture.png

iRobot also had to change its hiring practices to move to this new approach to tackling the smart home. "If you go back five years we had different software teams for different robots and it became difficult to hire new talent," Angle says. Typically algorithms were tuned to one product and the assembly language was custom.

"We had hundreds of software engineers that needed to standardize on a common language. Now we use Python and have common code bases and teams can move from one robot to another," Angle says. "We are moving from wonderful but old languages and software infrastructure to modular generic snippets of code."

The company has also put a big push on developing talent that can combine tools such as Python, Raspberry Pi, and robotics. iRobot, with its Create 2 Project, has a middle school curriculum for Python and developer kits to create a robot running via Raspberry Pi, cameras for sight, and PrimeSense for 3D sensing technology.

Also: Cracking Open the Roomba 980 robot vacuum (TechRepublic)

iRobot devices still run on a proprietary operating system, but Angle says the company will ultimately move to something more common like the emerging ROS 2 robot operating system. The catch is that today's robot operating systems need more processing power and are memory intensive. As the industry grows, Angle is betting that there will be a common operating system layer.

By paying down technology debt, iRobot can focus on what it does best. Angle considers the company's crown jewels to be AI, navigation algorithms and design capability.

"The role of iRobot in the emerging smart home is to be the trusted holder of spatial understanding so the home can be actually intelligent," Angle says. That strategy is fundamentally different than focusing on the voice platform or tying smart home devices together with a phone. "Organizing spatial data in the home is important and defendable. It is also enabled by a trusted relationship with the customer," Angle says.

According to Tirozzi, iRobot came to its spatial data focus based on a simple problem it was trying to solve: "The genesis of the spatial data was ensuring the robot can navigate. The goal was to be able to clean the kitchen as we worked backwards," says Tirozzi. "Like any agile organization you realize there is more value there. Spatial understanding allows consumers to view their space in a different way."

s9-s9-photo-insitu-imprint-smart-map.jpg

iRobot

Tirozzi added that spatial data is the key to personalizing experiences across different types of households. An empty nester will have a different cleaning routine than a couple with four children. 

Dorsheimer, an analyst at Canaccord Genuity, points to the launch of iRobot's Roomba i7 last year as a game changer for the company as the robot became more autonomous, and coupled with its Clean Base, more low maintenance. "This will be a core component as the company transitions from a technology-based company to a network orchestrator," says Dorsheimer.

Angle says the rollout of the i7 and i7+ gained traction in EMEA and Asia in the first quarter. "By analyzing the information available to us with permission of our customers, we're able to see their level of engagement with our connected products, how often and for how long they're using their Roomba, how frequently missions are completed, whether they are using directed room cleaning available in our latest premium model and so on," he says.
 
To Angle, the smart home isn't about convenience as much as it is a secure place that operates by itself with affordable technology.

Ultimately, the smart home is going to be enabled by trust. iRobot is GDPR compliant and will forgo monetization tied to customer data. "Improving the consumer experience is how we win. That's how more people buy our products. Data enables products to work better together," he says. That means iRobot's smart home strategy is very different to big tech companies in the space which are using it as an opportunity to acquire even more data about us all and monetize it.

iRobot has published its core privacy principles and for good reason. "Privacy is built into how we develop products," Tirozzi says. "You need trust to have someone invite our products into the home."

And that trust could be a good foundation as iRobot expands into new areas. iRobot appears to be stepping up its product cadence. Including the launch of Terra later this year, iRobot will have launched five new products in the last 12 months.

Download this article as a PDF (free registration required).

Building a data science team

Angle says iRobot's competitive advantage may ultimately be its data science team. Angle says there are 10 million robots in the field, and data science is how the company will optimize and improve operations. "Things like how are they getting stuck, what rooms, connecting robots has a huge impact on us, and data can tell us if there's a problem with the robot so customer service can rapidly diagnose," Angle says.

iRobot started building its data science team in the same place as many enterprises -- in information technology. Why? That's where the passion was for data science. Tirozzi's role has evolved as iRobot began to focus more on the data.

"We had an enterprise data warehouse, but not a platform. We found a few data scientists and proved some things quickly. At that point, data scientists were spending 80% of their time cleaning data and 20% on data science. Once we demonstrated that we could prove hypotheses and work with executives we moved up the chain to patterns and predictions," Tirozzi says . "Data science is a core part of our DNA."

The evolution of Tirozzi's role reflects how iRobot and Angle approach the data overall. "It became clear over time that the data science was improving in IT, but it was too far away from the product. Engineers weren't connected sufficiently. Last year we moved the data science team into engineering as a module," Angle explains.

Angle added data science is a function that cuts across operations, customer service, and products. The data science team at iRobot is designed to be partners with engineering and product design. "Data has an impact on all of our products," Angle says. iRobot's business and approach to data science has been detailed in the Harvard Business Review with management tips from Angela Bassa, director of data science at the company.

Indeed, an effort like iRobot's imprint mapping gives the robot memory. "If you have memory you can learn and that learning requires data and information as a foundation," Angle says. Another example where data science informs product design is the Roomba Clean Base.

"We need to know how many times Roomba can run before it gets stuck with Clean Base. It may be months if not a year before you touch the robot," Angle says. "Data science becomes central to product efficiency, design, and software parameters. If the robot doesn't make it back to the dock its life has ended."

Data science is critical to iRobot's future because Angle's vision revolves around morphing Roomba from a consumer product to more of a service.

Download this article as a PDF (free registration required).

Culturally, iRobot found early wins on data science as well as internal advocates and new talent. Once teams start asking questions about products and receive data to drive insights, they tend to get on the data science bandwagon, Angle says. This integration of data science into functional teams ultimately changes culture.

"Cultural change is why we are spending money on data science," Angle says. "iRobot is transitioning from a largely hardware company to one about data."

How far can iRobot go?

The big unknown for iRobot is how far it can go with its approach to the smart home. Amazon and Google have already extended their business models -- e-commerce and advertising -- into the smart home. Amazon has used Alexa to serve as a smart home hub and connect to various things in the home including iRobot devices. Google Home's strategy revolves around information with a dash of e-commerce, advertising and Google Assistant. iRobot has partnered with Google, too.

Other tech giants such as Apple and Samsung aim to connect their footprint in the smart home and connectivity providers such as Comcast also are aiming to be the central nervous system in the home. Many of these tech giants are looking at the smart home as a place where smart speakers and smartphones are the remote control for the home. iRobot partner Amazon may also have its own home robot ambitions.

Add it up and iRobot could ultimately become an acquisition for a tech giant, but the reality is that the company may turn out to be a better partner for now.

Tirozzi chuckles when iRobot's almost 30 year history is mentioned. "It doesn't feel like we're 30 years old. We're a midsized company with a startup culture," he says. "We're a bunch of nerds at heart and the sky is the limit. We're at the forefront of this and looking at how we can impact lives. How much friction can we reduce? How do we help people live in place? We see iRobot as a cornerstone of the smart home. As Colin says, we're just about none of the way there."

m6-silver-photo-studio-frontfacinghero-singleuse-wetpad-phone.jpg

iRobot's Braava Jet M6 mopping robot

iRobot


For now, iRobot's expansion strategy revolves around global expansion. The company has household penetration of 11% in the US and single digits globally. The plan is to grow its customer base and integrate an ecosystem of connected robots.

However, there is a bevy of competitors that have more resources than iRobot. In a regulatory filing, iRobot lists consumer electronics giants such as Samsung, LG, Xiaomi, Cecotec, and Shark as competitors with traditional floor cleaning brands such as Dyson, Bissell, Hoover, and robotic cleaning outfits like Ecovacs and iLife.

In the end, iRobot's smart home bet is that navigating your home will put it in the middle of the ecosystem with data science enabling personalized routines. In a regulatory filing, iRobot noted that it plans to continue to "identify additional ways to advance the smart home experience by enabling a broader understanding of the home's space, enabled through Roomba's spatial awareness of the home."

Or, as Angle puts it: "If a robot is going to get a beer out of the kitchen, it needs to know where the kitchen is."

Download this article as a PDF (free registration required).