Colin Ingres, CEO of iRobot: Lidar is a dead end, and the future home is intelligent robots

The Heart of the Machine is original

Author: Wen Fei

iRobot, the pioneer of sweeping robots, has temporarily performed poorly in the Chinese market, but CEO Colin Ingres has its own technical route and strategic planning: he firmly believes that only vision-based, collaborative intelligence is the ideal form of home service robots; the future house or "home" will become intelligent robots, and iRobot will occupy a place in the core of the smart home ecosystem with visual and environmental understanding and autonomous mobility capabilities.

At first glance, this seems to be another case of localization failure - in the field of sweeping robots, iRobot, the world's largest market share, in Chinese consumers, the popularity of even entering the top three needs to be squeezed.

Looking at this, this seems to be a competition of capital to develop the imagination of the company - the valuation of iRobot in the 31-year-old wall Street is less than one-third of the 7-year-old Stone Technology on the Shanghai Stock Exchange.

However, in the view of Colin Angle, CEO of iRobot, these, and other troughs, are just a bit of a landscape that will be experienced on the way to climb the highest peak of the intelligent service robot swarm.

The roboticist, who also developed a non-investment, no-financing MIT start-up team into a public company that created and defined the commodity category of "household cleaning robots", firmly believes that only intelligence based on visual understanding and can cooperate with people tacitly is the ideal form of home service robots, and iRobot is fully prepared for this.

From the initial ant-inspired space exploration robot that "can work without thinking like a person", Colin led iRobot to develop more than 100 robots, involving more than 20 industries such as aerospace, deep-sea, security, medical, toys, etc. These products have different levels of intelligence and different shapes and sizes, but the purpose is consistent, that is, it must be practical, and the value provided must be higher than its manufacturing (or sales) cost.

In the intelligence system built by iRobot, intelligence is hierarchical, and all underlying behaviors do not require planning except for a small part of the topmost layer, similar to the subconscious response of humans. "By focusing first on building the underlying behavior, and turning as much intelligence into behavior as possible, top-level planning can be simple enough to really work," Colin told Machine Heart. It was this different understanding of machine intelligence, and the way intelligence was built from the bottom up, that laid the foundation for iRobot.

Although the advent of the intelligent era made Colin say that he needed to rethink the definition of "robot" - from "tool" "helper" to "partner", can perceive the user's intentions and needs, and build a certain emotional connection with humans, but can feel what he called "emotional connection", still able to let people use intelligent robots without emotional burden, just like you and I are now at ease 7× 24 hours to use smart phones. "This is the service robot that the aging society will need in the future," Colin said, "and making humanoid robots has wasted too much time, money and energy."

Transforming from an initial defense contractor to today's small home appliance giant, Colin now aims to lead iRobot from a successful robotics company to a company that can completely solve the concept of "understanding home", and thus occupy a place at the core of the future smart home. He believes that the human habitation pattern has not been considered from the perspective of intelligence, in his vision, the future of the house itself will become an intelligent robot, indoor and outdoor various networked IoT devices, just like the "home" of this robot component, and the wandering home cleaning robot Roomba, will rely on visual understanding and autonomous movement and other capabilities, like cleaning up germs and garbage white blood cells in the human body, become an important part of the smart home ecosystem.

In the interview, Colin repeatedly mentioned Collaborative Intelligence's "collaborative intelligence", reminiscent of His mentor and business partner Rodney Brooks, who founded Rethink Robotics, which pioneered the new product category and technology field of "industrial collaborative robots" with sawyer and Baxter robots, but in 2018 He stopped operating and sold his assets, in Colin's words, "forgetting to be a great company on the way to becoming a great innovator."

In the increasingly competitive home service robot market, can iRobot accurately gain insight into user needs, just as it used Roomba to define the sweeping robot, take into account technological innovation and business expansion, and lead the future smart home with better products and services?

Colin Angle, CEO of iRobot, and the company represents the latest roomba S9+ sweeping robot

Lidar is a dead end and doesn't waste time on "harmful" technology

At present, a big reason why iRobot has encountered bottlenecks in the Chinese market is that the product adopts a pure visual navigation scheme and does not use any laser sensors. From the perspective of consumers' use experience, it is clear that the initial drawing time of the iRobot sweeping robot will be slightly inferior to that of competitors using lidar.

Commercially available high-end household sweeping /wiping robots, there are three main types of navigation technology, one is laser navigation, one is visual navigation, and one is laser + visual navigation. From the naming method alone, it can be roughly inferred that, at least at this stage, laser + visual navigation must be a combination of the advantages of laser navigation and visual navigation, and this is indeed the solution used in the current hot sweeping robot products.

Laser navigation and visual navigation are two different navigation technologies, and the industry is inconclusive about the above two technical routes. However, there is a good reference case for the pure visual route and the lidar dispute - Tesla's self-driving development plan. Musk has said more than once in public that lidar has no doomed, and Tesla's multiple fatal accidents have not eased his tone.

In the use of lidar, Colin and Musk heroes see a little bit the same. "We developed a product with lidar 25 years ago and know it's a dead end. ...... Lidar may be useful in the short term, but to be truly intelligent, the camera is the right sensor," Colin said.

He was well aware of the controversial nature of the decision. But for Colin, lidar not only bypasses the problems that must be solved to achieve true intelligence, but also poses a hidden danger like a buffer overflow, which will only consume more time and effort in order to eliminate this hidden danger in the future. So adding LiDAR to Roomba is simple, but he will never invest in technology that he knows is "harmful."

"Maybe lidar will make me sure there's something out there, but if I don't solve the visual comprehension problem, I can't know what that thing is." "Any reduction in understanding of the environment is a distraction," he stressed.

Colin previously said in an interview with foreign media that iRobot is the company with the second largest networked robot product in the world after Tesla. In 2020, iRobot launched the iRobot Genius home intelligence platform, which can be seen as a further upgrade of the unified management of tens of millions of cleaning robots distributed around the world, and what a Roomba "learns" in a home will be shared by each Roomba, which is a continuous iteration of continuous development of collective intelligence.

The data that iRobot has collected and accumulated over the years is undoubtedly a powerful asset and one of the reasons for Colin's confidence, "We have the smartest robots in the world," he said. Like Tesla, every few months, iRobot upgrades its software to add new features to the product. But unlike Tesla, roomba's absence of lidar is not "deadly", and iRobot can safely take the pure visual route.

Turn 90% intelligence into behavior and build "collaborative intelligence" from the bottom up

In the field of robotics, the challenge of developing a human-level cognitive computing model is to make a model that is both accurate and versatile, but unfortunately these two goals have always been contradictory, because to build a more realistic model of rational behavior, it is necessary to consider the cost of calculation.

The key to iRobot's initial success was to follow the philosophy that robots can work without thinking like humans, using the so-called Subsumption Architecture. Unlike the prevailing view at the time, this architecture did not guide the robot's behavior by constructing a complete digital representation of the external world, but by decomposing the behavior into a series of sub-behaviors, so that the robot could interact and respond to its surroundings in real time. These sub-behaviors are organized into a hierarchical structure, each layer can achieve a specific level of behavioral ability, and the combination of low-level behaviors can achieve higher-level behavior. The first generation of Roomba was born out of this, with basically no planning in mind.

Colin believes that this bottom-up approach to building intelligence is the foundation of iRobot, and that it can be extended to building human-level intelligence. He explains: "Intelligence can be seen as a combination of low-level behavior and high-level thinking based on planning. If I can turn 90% of my intelligence into behavior, then I just have to plan in that thin layer above. ...... How much can I do with low-level behavior? The more I do at the lower levels, the simpler the hard parts will be."

The difficult part is building "collaborative intelligence", which is also Colin's ultimate goal to build service robots. This intelligence allows robots and people to cooperate and cooperate tacitly, thus building trust, so that people can rest assured that robots can do more and more complex things.

Today, more and more robotics researchers believe that it is necessary for robots to think rationally about humans – not just as obstacles or perfect gamers, we need robots to take human nature into account so that they can better cooperate with humans, that is, to achieve so-called "value alignment". If successful, we will have the tools to dramatically improve the quality of life.

Achieving behavioral compatibility requires robots to predict human behavior and understand how these human behaviors will affect their own behavior, while also enabling humans to predict the next behavior of robots. In addressing these challenges, Colin said iRobot has made some progress, but there is still a long way to go. "We're at the forefront of the intelligence we need to build a smart home, and it's very exciting... Every day we try to figure out what's holding consumers and robots back from forging closer partnerships, how to build trust, how robots can respect and empathize with consumers' home habits, and how to foster similar emotional connections between consumers and other devices in their homes."

"In order to build a collaborative partnership, you need to understand what you're working on to maintain. In a way, this is not a breakthrough in artificial intelligence, but a breakthrough in the learning and understanding of the environment . It goes back to the question of why I don't like lasers," Colin says, "because I want to build the richest environment to build my intelligence, and lasers don't help with that."

In his 32 years in robotics, Colin has seen many different problems, some of which are definitely best solved through hardware solutions, but for today's sweeping robots, he believes that the main challenge comes from intelligence, not cleaning efficiency during the movement. "I don't care that iRobot's sweeping robot spends a few more minutes mapping, I want the robot to really understand the home it cleans."

Building a robot is the ultimate test of system design, requiring great care to understand and define the problem, choosing the right combination of hardware and software. Hardware adds cost, fragility on the system, and unnecessary complexity, while software is not only upgraded regularly, but almost always cheaper. As a result, iRobot will continue to build Colin's ideal machine intelligence around vision, which he believes applies not only to Roomba, but also to broader robots and larger smart homes.

In the future, the "home" will be an invisible intelligent robot

On the path to the end goal, Colin was very careful not to get stuck in a local minimum. He says iRobot thinks not just about how to deal with the simplest answers to the question of the present, but about building the best answers to today's problems that can effectively extend to tomorrow.

Over the next 20 years or more, he will lead iRobot in his efforts to address the concept of the "understanding of the intellect." His ambitions are obvious, because it's about human life, something that no company that puts AI first or smart technology as its core selling point can ignore.

When it comes to smart homes, many people don't think deeply about the meaning of the word. Ask a person with more than 20 smart devices in their home, is your home smart? Usually the other party will say that they are not smart. Colin pointed out that this is because consumers don't really know what intelligence is, but if you ask another question: "Do you want your home to be a clean, orderly, safe, environmentally friendly, comfortable and safe place?" I believe the answer will be, "Yes, I'd love to pay for such a change."

Colin believes that the "home" will eventually become invisible intelligent robots, or at least function as a system. The home here includes residences, living, indoor and outdoor activity spaces, etc., it is very intelligent, its task is to take care of and protect the people living in it.

"Amazon or Google may be present in future smart home systems, but iRobot or a company similar to iRobot is still needed to perfect it." Colin laughs and says, "I don't need to be the center, but without us, the human family can't be intelligent."

Nowadays, many countries and regions, including China, are rapidly entering an aging society. What kind of home service robots do we need in the future? For this problem, Colin believes that humanoid robots are not the solution.

"The ideal shape for a home service robot is one that can accomplish its assigned tasks in the most reliable, safe and efficient way, whereas humanoid robots are too expensive and too risky." Colin said. In the future, there may be many different robots in the home that solve specific problems, and these robots can work together long before we have the ability to build humanoid robots. Aside from entertainment, such as serving visitors at Disneyland, he had a hard time finding a scene where humanoid robots could come in handy.

"The solution to most everyday problems won't be to build a humanoid robot and let it operate a device or instrument originally designed for humans." Colin bluntly said: "Trying to build humanoid robots has wasted a lot of time, money and energy."

Colin says a lot of companies make great robots, but they can't find a reason to do so. Boston Dynamics' work is stunning, "but these robots don't have real value yet." If I were the CEO of Boston Dynamics, I would build a pet robot, and I would use all these amazing dynamic controls and jumping techniques to make a puppy that could run over and jump on my lap and I could stroke it." "The most valuable application of Boston Dynamics' technology could be to create robotic pets to solve the problem of loneliness," Colin said.

He thinks Google is also a bit too focused on innovation and spends too much money on moon landing projects. Colin stressed that how to balance technological innovation and business operation is something that companies, especially technology startups, must always keep in mind.

Sticking to cutting-edge technological innovation premised on pragmatism has allowed iRobot to survive for 31 years and become a name to be reckoned with in the global robotics industry. Asked how he would like to be remembered, Colin does not hesitate to say that he considers himself a Builder "builder" and that he hopes to be remembered for building a practical robotics industry.

In Chinese context, Builder feels more like a "pit filler," building things to solve problems. For Colin, Colin said, iRobot wants to be a company that builds collaborative intelligence and creates robots that can help humans live independently for longer, which will be a lifelong journey for him. In this sense, the company's latest slogan does convey his heart:

iRobot, so you can human.