Qin Baoxing of Gaoxian Robotics: 2019 will be the year of mass production of service robots

In 2019, Internet giants were making structural adjustments, autonomous driving was experiencing an industry winter, and service robots began to speed up mass production.

The first two are already an industry consensus, and this article will talk about the latter, service robots in 2019.

On January 16, Yogo Robot, a delivery robot company, released YOGO Station and announced that related products will be delivered in batches by the end of the year.

On February 18, service robot company Yunji Technology announced the completion of its Series B financing, deepening its layout in the smart hotel market;

On March 19, service robot company Gaoxian Robot announced the completion of its Series B financing, with the funds mainly used to achieve mass production and delivery of products.

"In fact, our products are already mature, and now we need to release production capacity as soon as possible. This is also one of our work priorities this year." Qin Baoxing, CTO of Gaoxian Robotics, told Leifeng.com.

However, before talking about mass production, we must first understand the current basic status of service robots in terms of technology, products and market. Leifeng.com discussed these issues with Qin Baoxing, CTO of Gaoxian Robotics.

Service robots can be divided into two categories: B-end applications and C-end applications (consumer robots). Among them, sweeping robots are typical C-end applications.

Due to different application scenarios, the specific scenario requirements are slightly different. Taking the composition and positioning functions of sweeping robots as an example, sweeping robots are currently mainly used in home scenes, where the lighting changes are not very strong and the consequences of positioning loss are not very serious. Qin Baoxing told Leifeng.com that the main differences between commercial (to B) service robots and civilian (to C) service robots are as follows:

• Commercial service robots have a larger operating range and are more dynamic in terms of application environment, such as mapping area and positioning accuracy;

• The operating environment of commercial service robots is more dynamic, requiring robots to have stronger perception and path planning capabilities;

• Commercial service robots have higher requirements for system maturity and robustness.

In short, consumer robots emphasize functionality more, while service robots must first ensure accuracy.

As one of the key systems of a robot, should the navigation system use machine vision or lidar?

At this stage, laser radar applications are still the main focus in China, especially in the commercial field.

Qin Baoxing told Leifeng.com that the system built with machine vision is not very robust and cannot meet the needs of large-scale (B-side) commercial use at this stage. At present, machine vision is still mainly used in relatively structured scenes with little light changes or relatively simple scenes such as homes.

LiDAR technology is relatively mature nowadays, and because it is not affected by light intensity, it is more suitable for positioning and navigation. "But machine vision is indeed a major trend at the moment, and everyone is studying how to enhance its robustness. In the long run, the two will inevitably be combined for application."

How will the two be combined and applied in the end?

LiDAR is used to provide robust geometric information for the positioning system, and machine vision is used to provide some advanced semantic information.

Qin Baoxing explained it using the example of a high-speed railway station with large passenger flow.

"There are many people in the high-speed railway station, so there are too many feature points that are not fixed. In such an environment, it is difficult to use only lidar for positioning and navigation. The semantic information provided by machine vision can identify which are moving people and which are fixed feature points (such as pillars). This provides semantic information for mapping and positioning. Combined with the application of lidar, accurate positioning, composition and navigation can be achieved."

It can be seen that the future application of robots in navigation, positioning and composition will be a trend of multi-sensor fusion.

In fact, Gaoxian Robotics not only applies multi-sensor fusion technology to mapping and positioning, but also to perception.

In terms of mapping and positioning, Gaoxian Robotics first uses LiDAR for SLAM navigation, and then uses machine vision to generate some closed-loop information. "Such fusion applications have been used in our products for more than a year," Qin Baoxing told Leifeng.com.

In terms of perception, Gaoxian Robotics has also begun to use machine vision to extract semantic information in some of its products. Taking the patrol robots previously used in high-speed rail stations as an example, "What Gaoxian Robotics is doing now is to distinguish pedestrians from pillars through deep learning, and then match them with the information obtained by the laser radar to identify which laser points correspond to fixed points and which correspond to unfixed points, and obtain the optimal position estimate through algorithms to improve the robustness of our robot positioning system."

Qin Baoxing told Leifeng.com that Gaoxian Robotics has three main functional developments in mapping and positioning this year:

First, "unlimited" maps . "People often ask us how big a map we can build. In fact, our initial map-building capabilities were often limited to hundreds of thousands or millions of square meters. Our goal this year is to no longer be limited by area. In theory, the size of our hard drive determines the size of the map we can build." This capability will make it possible for Gaoxian Robotics to apply related products in large-scale shopping malls or municipal roads.

Second, apply semantic mapping and positioning to all product lines. "We already have some small-scale product lines using it, and this year we will expand this capability to our B-side customer products or solutions."

Third, real-time map updates.

Qin Baoxing told Leifeng.com that perception can be divided into two types of applications:

First, perception related to autonomous mobility . In specific application scenarios, unmanned or autonomous mobile robots need to be able to identify static obstacles and dynamic obstacles, and be able to identify and classify and track them, so as to make behavior predictions more convenient. Gaoxian's existing cleaning robots can ensure the safety of pedestrians on the one hand, and bring better interactive capabilities on the other hand by adding human detection and face recognition technology;

Second, business-related perception. In specific application scenarios, the cleaning robot must first identify the floor in the scene, such as carpet, marble, or cement, to determine which cleaning mode to use. This also requires the cleaning robot to have the ability to perceive the environment.

Qin Baoxing told Leifeng.com, "Our technical thinking is: we base our technology on the perception related to autonomous mobility and make it solid and practical; at the same time, we continue to explore the perception capabilities related to the business and provide real value to users."

Whether it is infinite maps, semantic maps, real-time maps, or the update and iteration of perception capabilities, it will consume a lot of computing resources. Generally speaking, there are two types of solutions:

First, upgrade the hardware, and upgrade and iterate the hardware performance and hardware configuration;

Second, update the software, optimize the software algorithm and iterate the version.

Among them, the former is relatively simple and direct, which will lead to an increase in hardware costs and even the complexity of hardware resources; the latter is more complex, requires higher investment in human resources, and may also have a certain impact on the R&D cycle.

Gaoxian Robot’s choice is the latter.

The premise of our research and development is to upgrade these functions by optimizing the algorithms and iterating the software versions without changing the hardware.

"We are actually quite special in one aspect: we have the characteristics of an Internet company, which pursues rapid iteration, and we also have the characteristics of a robotics company, which pursues product stability and maturity," Qin Baoxing told Leifeng.com.

These two seemingly contradictory points are achieved by Gaoxian Robotics through both software and hardware, namely, rapid iteration of software and algorithms and pursuit of hardware product stability. Qin Baoxing revealed:

In terms of software, Gaoxian Robotics makes a major version update once a month;

In terms of hardware, Gaoxian Robotics usually iterates its self-developed products two or three times a year.

Qin Baoxing specifically explained the hardware iteration: Because B-end products involve many supply chain issues such as mechanical structure, mechanical processing, and mold opening, as well as the market acceptance and market education issues of new products, we will not make frequent iterations on hardware.

Leifeng.com also agrees with this point. The product iteration strategies of B-end products and C-end products are different. The rapid iteration of C-end products can promote consumption, while B-end products pursue stability more. The iteration speed will not only bring pressure to B-end customers, but also bring educational difficulties, such as the cleaning lady is not willing to learn the new version of the cleaning robot mobile APP.

Qin Baoxing told Leifeng.com that even when updating software functions like mobile apps, Gaoxian Robot will try to be simple and easy to use.

Service robots in some industry applications are now criticized. For example, in banking scenarios, the large touch screens currently used can actually meet most functional requirements, and mobile robots can bring more entertainment value. What more targeted and rigid functions do robots that need to be used in mobile scenarios, such as wheels or even feet, need to provide? Further exploration is needed.

We take business needs into consideration during product development and provide our products and solutions for scenarios that truly require robotic mobile operations.

In this regard, Gaoxian Robotics has selected three application scenarios: cleaning (commercial), security and building delivery.

Among them, in the field of commercial cleaning, Gaoxian Robotics has the Aikebao series of robots, which mainly output products; in the field of security and building distribution, it mainly outputs solutions and provides security and building distribution solutions to other manufacturers.

Qin Baoxing particularly emphasized that Gaoxian Robot does not provide robot solutions in the vertical field of commercial cleaning (only products). This is also Gaoxian Robot's conscious avoidance of possible conflicts when providing both product and solution business models.

Why choose to take into account both products and solutions?

Qin Baoxing believes that "one advantage of making plans is that you can get in touch with user needs at the first time."

When we make products, we are actually not sensitive to many user needs. When we make solutions, we will encounter various cooperation plans and various needs. In this process, we continuously improve the robustness of our algorithms and the richness of the overall solutions by solving these needs, so as to improve our technology. These technologies, in turn, first empower customers and can also be imported into our own products.

It can be seen that Gaoxian Robot regards providing solutions as growth, honing its technology by providing solutions, and empowering new solutions and products.

Since its inception, Gaoxian Robotics has mainly focused on overseas markets including seven countries including Singapore. In 2015, the first batch of cleaning robots landed in Sentosa, Singapore.

It is precisely with these two legs that Gaoxian Robot began to move from abroad to domestic market.

China has a large population base. At the same time, as the problem of aging population gradually emerges, there will be a large labor gap. We definitely need robots to fill these gaps. Therefore, there is a strong demand for robot products in China. This is also one of the reasons why the domestic robot industry is booming, a large number of robot manufacturers have emerged, and even foreign robot manufacturers are gradually moving to China.

Since the second half of last year, our business center has shifted to China. At present, our cleaning robots have a market share of more than 90% in the domestic market, and the large-scale mass production of cleaning robots is also our focus this year.

When everyone is talking about 5G, what can 5G bring and what can it bring to the robotics industry?

In particular, the development speed of domestic 5G is much higher than expected. Therefore, this is also a technical field that every product manufacturer and service manufacturer needs to consider and plan in advance.

Qin Baoxing told Leifeng.com that for Gaoxian Robotics, the main impacts are as follows:

• First, data transmission. Previously, due to bandwidth constraints, robots could not transmit all real-time data back to the background. Our current robots upload data when they return to the workstation or charging station at night after running for a day. After 5G becomes popular, we can achieve real-time data transmission, so people can monitor the working status of robots in real time.

• Second, remote control. When robots are deployed on a large scale, a central control platform is needed to realize remote control of the robots. 5G's low latency can meet such work needs.

• Third, cloud computing. The high bandwidth and low latency brought by 5G naturally allow most of the computing to be placed in the cloud, but security issues need to be considered here. Placing it in the cloud will naturally lead to the problem of over-reliance on transmission. When your backend is hacked, the entire security will be affected. However, whether this function should be implemented remains to be further explored.

Specifically for the development of 5G, Gaoxian Robotics has already laid out the first two points: data return and remote control. In addition, Gaoxian Robotics is currently implementing perception computing and perception reasoning on the device side.

On March 19, at the Gaoxian Robot 2019 Media Communication Conference, the company announced its 100 million yuan Series B financing, including the introduction of a new investor, Far East Horizon Group's Yuanyi Investment, and the joint development of an equipment leasing business model in the property management field.

Qin Baoxing told Leifeng.com that Gaoxian Robotics' Series B financing will be mainly used in three aspects:

• First, dig deeper into the technology, including working with universities on research and development;

In terms of cooperation with universities, Gaoxian Robotics also mentioned at the press conference that it will carry out in-depth cooperation with Tsinghua University, Shanghai Jiaotong University and the National University of Singapore, and will build three R&D centers in Beijing, Shanghai and Singapore starting from 2019. Qin Baoxing explained:

We have reached a three-year cooperation project with Tsinghua University (starting in February this year), mainly using deep learning to solve the indoor and outdoor perception problems of autonomous driving;

Collaboration with Shanghai Jiao Tong University mainly on robot sensors;

The collaboration with the National University of Singapore is mainly on building positioning.

• Second, introduce core talents, including attracting elite talents from mature robotics industries and other industries;

• Third, mass production of products, "which is also the focus of our financing output."

In addition, Qin Baoxing told Leifeng.com that in 2019, Gaoxian Robotics' main tasks will be two aspects: mass production and standard solutions.

First, mass production of cleaning robots. "Our products are mature, and we need to release production capacity as soon as possible. This is our financing focus this year and also our work focus."

At the same time, we will also specifically address technical issues such as robot stability and maintainability after large-scale mass production, and continue to enhance our capabilities on the robot platform to support our mass production plans.

Second, standardized solutions. "Last year, we focused on robotic solutions and developed many customized solutions. This year, we will develop standardized solutions and provide robotic solutions as standard products to our partners."

In addition, Qin Baoxing also revealed to Leifeng.com that for next year, we are actually actively promoting the implementation of unmanned driving solutions and unmanned driving vehicles.

Service robots are like a group of children, going through several stages of childishness, rebellion and maturity in different vertical fields. From entertainment to functionality, from making up for the differentiated needs of the industry to providing the rigid needs of the industry, each field has a different positioning and will also produce robots of different forms.

At the same time, as a highly complex intelligent system, manufacturers are still exploring how to bring out its true value in various fields, bring about industry innovation or improve industrial efficiency.

However, as many robot manufacturers have announced that they will enter the mass production stage, 2019 will be another year of great pressure for robot manufacturers amid everyone's expectations.

Tencent responds to allegations of unfair competition; 40% of AI companies are fake AI; Jia Yueting sells Faraday Future headquarters

Google Cloud Gaming is here: You can play 3A masterpieces in your browser, cloud gaming is about to subvert the future!

Tencent refutes the "Duoshan App pop-up window" incident; Apple responds that iPhone 6 will be discontinued in May this year; Xiaomi IoT platform connects 151 million devices

GTC 2019: No new architecture, no big nuclear bomb, Huang Renxun built a huge circle of friends

The United States launches 6G research and development; Australia approves Huawei to provide digital communication systems; Lyft launches IPO roadshow

Follow Leiphone.com (leiphone-sz) Reply 2 Add readers group to make friends