1.The embarrassment of intelligent security robots in reality

Guide: The landing and application of AI+ robots in security scenarios is a promising future.

At the 2019 World Mobile Communications Exhibition (MWC2019), Siasun and Liaoning Mobile released the first demonstration application project of the "5G Smart Factory Application Demonstration Project" - "5G Intelligent Inspection Robot", which has entered the actual network testing stage. In the near future, it will be applied to industrial enterprise parks such as Brilliance, BMW, Xinsong and Liaoning Machine Tool to assist in security work.

The application of 5G+artificial intelligence + security + robots has once again set off an AI wave in the industry. Looking back at 2017, the China Academy of Information and Communications Technology released data showing that the number of artificial intelligence patent applications reached more than 300,000 that year, and the total investment and financing scale reached 180 billion yuan, which once made artificial intelligence the biggest hot spot in 2017.

Robots have been used in the security industry for many years. In 2014, CIIC developed the first practical security patrol robot in my country, bringing security robots from concept to practice.

The continuous development of the past five years has also given new attributes to security robots, such as AI, 5G, face recognition, etc. Although the situation is good, it is difficult to hide the embarrassment of intelligent security robots being "not enough compared to the top, but more than the bottom" in practical applications. Undoubtedly, within 5 years, the technology of security robots has been continuously optimized and their performance has been upgraded. In particular, with the rise of the "robot +" trend in recent years, more and more companies and capital have invested in it, accelerating its development. The theory of intelligent relativity found that compared with the first domestic security patrol robot, the current intelligent security robot has made significant progress.

With the update and iteration of technology, security robots are making progress.

Undoubtedly, within 5 years, the technology of security robots has been continuously optimized and their performance has been upgraded. In particular, with the rise of the "robot +" trend in recent years, more and more companies and capital have invested in it, accelerating its development. The theory of intelligent relativity found that compared with the first domestic security patrol robot, the current intelligent security robot has made significant progress.

One, all aspects of audio and video surveillance upgrade. The security patrol robot released by CIIC is equipped with a 360° panoramic high-definition infrared integrated night-vision pan-tilt camera and standard-definition cameras around the body, which can stabilize video surveillance images in various complex environments day and night, and is equipped with a pickup to collect live sounds in real time, which can provide real and accurate on-site real-time conditions for the security background. The ATRIS series robots released by UBTECH Robotics led by Tencent are specially customized for security patrols. They are equipped with thermal imaging dual-spectrum network light-load pan-tilt cameras. Even in low light and rainy weather, the infrared imaging principle can be used to accurately identify the field of view. In terms of audio and video surveillance, intelligent security robots across the country are constantly being upgraded, raising the average level of the industry.

Two, super endurance guarantee. The "5G Intelligent Patrol Robot" released by Siasun is equipped with a lithium battery that can last for more than 8 hours at full load, and has the ability to independently detect power and charge. Coincidentally, the intelligent security robot ANBOT released by Wanwei also enters the security market with the ability of 8 hours of battery life and 2 hours of independent charging. This is the normal state of the machine. When people are still looking for a balance between work and life in the discussion of 996, the robot has reached the 24-hour continuous working state of 007, making up for the time and position of artificial absence.

Three, more intelligent performance. In the early days, the K5 patrol robot developed by the American Knightscope company made a lot of jokes. Although the robot itself is equipped with smart devices such as cameras, GPS positioning and sensors, it has repeatedly committed suicide by jumping into the pool and colliding with children. People are worrying. The technological progress and algorithm upgrade of AI endow robots with more intelligent behavior performance. The "e-patrol" machine police released by Avenue Zhichuang has more accurate feedback in face recognition, license plate recognition and voice recognition, VSLAM positioning and environment reconstruction.The function makes the real world in the eyes of the robot appear digital, and improves the reaction speed of the robot in the real environment.

With the development of China's economy, the flow of personnel between regions has become more and more frequent, which has gradually increased the pressure on security. The intelligent security robot, which is at the forefront of science, is used in major public places while improving its performance. At present, Wanwei Robot's intelligent security robot series products are serving Guangzhou Baiyun Airport, Shenzhen Bao'an Airport, Inner Mongolia Yili Park, Saudi Arabia King Abdullah University of Science and Technology, etc., and the social feedback is good.

So, can it be declared that the era of robots in the security industry has arrived?

There are many minefields in security, and intelligent security robots are " worse off than some "

The security industry has always been a very delicate existence in society. It involves multiple scenarios and applications of social life, including patrolling, monitoring, investigation, anti-riot, alarm, control, processing, etc., and has always been on the fringes of public opinion. To put it bluntly, when there are no problems, everyone is happy. Once a problem arises and cannot be controlled and dealt with in a timely manner, it will fall into a storm of public opinion and cause serious negative effects.

When entrusting security to intelligent security robots, do you consider whether the functions of the robot can meet the security needs? Insufficient than Shang, does "Shang" refer to human beings? No, what the theory of intelligent relativity wants to express here is the "upper limit", that is, the upper limit of the security industry, the most stable social state achieved with the assistance of intelligent security robots. Do the smart security robots currently on the market meet this requirement?

First, the focus of security is not to find problems, but to solve them.
When a dangerous factor appears, whether the intelligent security robot can be checked or solved in time is the key to discussing whether a security robot meets the industry standard. Given that security is a multi-scenario industry, it is difficult to unify standards. Taking the most intuitive anti-riot as an example, when a thug with a knife appears in a public place, can the current intelligent security robot guarantee the safety of the public?

In terms of safety execution, CIIC's security patrol robots can currently support media functions such as sound and light alarms, one-key SOS alarms, and two-way voice intercoms. This is obvious and embarrassing. In the face of danger, the intelligent security robot seems to only act as an alarm. From this point of view, it is far below the expectations of the public.

When the K5 robot of Knightscope in California was released in 2014, the media repeatedly emphasized a word to look forward to the future of intelligent security robots-"robot police". "Heroic" robots are upright and fearless. Most importantly, they have the ability to subdue violence and protect the safety of the people, becoming the "sacred needle" for social security.

However, judging from the functions currently displayed in the industry, the service focus of security robots will fall on monitoring and patrolling. Through comprehensive patrolling and monitoring back and forth in the scene, data in the real world is collected, and then analyzed and screened by AI algorithms. Investigate the risk factors that affect social security, feed back to the background, remind the manual or contact the police for processing. Therefore, the current intelligent security robot is not enough to become a "robot police" on its own.

It is an alarm calling "superheroes", and the real superheroes are still front-line security personnel and armed police. Second, the AI technology that the security industry relies on most is difficult to break through, and the actual performance cannot meet the expected goals. Today's intelligent security robots are doing nothing more than surveillance, and intelligent surveillance is superior to traditional surveillance in the past because AI has intelligent functions such as facial recognition, voice recognition, and odor recognition. Taking the most commonly used facial recognition as an example, today's monitoring is not only required to "see" a certain person, but also to "know" this person, record his data and behavior in the cloud, and feed back to the background. This requires AI to accurately scan human facial features for recognition and judgment, but in the current market feedback, this technology is not as accurate as stated.

As early as 2017, when Apple released the iPhoneX, some foreign media invited the twins to test their Face ID, but the results were unsatisfactory. Until this year, there were similar news reports in China that twin brothers in Heilongjiang could still unlock each other's iPhone even though their faces were 70% similar, which is not only worrying.

Second, the AI technology that the security industry relies on most is difficult to break through, and the actual performance cannot meet the expected goals. Today's intelligent security robots are doing nothing more than surveillance, and intelligent surveillance is superior to traditional surveillance in the past because AI has intelligent functions such as facial recognition, voice recognition, and odor recognition. Taking the most commonly used facial recognition as an example, today's monitoring is not only required to "see" a certain person, but also to "know" this person, record his data and behavior in the cloud, and feed back to the background. This requires AI to accurately scan human facial features for recognition and judgment, but in the current market feedback, this technology is not as accurate as stated.

As early as 2017, when Apple released the iPhoneX, some foreign media invited the twins to test their Face ID, but the results were unsatisfactory. Until this year, there were similar news reports in China that twin brothers in Heilongjiang could still unlock each other's iPhone even though their faces were 70% similar, which is not only worrying.

When this technology is to be applied to security robots or the security industry, it is a social event worth pondering. Although there are mistakes in the mobile application, it may be just an individual problem in user experience or personal privacy. But once it is put into the public security industry that connects tens of millions of people, if AI cannot identify or identify the wrong person, will it bring unpredictable consequences? Does charging the wrong inmates allow the mob to get away with it? Or to distract the attention of the police and security personnel, etc.

Furthermore, there is a difference between "passive" and "active" individual behaviors in security facial recognition and mobile phone applications. When we use facial recognition on mobile phones, in order to improve the accuracy, users will consciously and unreservedly show their faces to AI for scanning. But in a public environment, ordinary people will not have this active behavior. Wearing masks, sunglasses or hats at the same time may affect the judgment of AI, which greatly increases the difficulty of AI recognition.

Of course, it would be a bit of a hooligan to single out facial recognition for discussion. After all, an intelligent security robot also has voice recognition, odor recognition, and even body recognition, etc. The launch of Galaxy Waterdrop has made the public see that the scope of AI recognition is expanding. Similarly, although the recognition capabilities of AI are being upgraded, it cannot be ignored that two major problems in the security industry will still hinder the accuracy of recognition. One is the complexity of the group, the flow of people is large, the mobility is strong, and the environment is complex; the other is the change of individuals, changes in clothing and appearance, and there are temporal and spatial differences.

Third, the industry market is chaotic, and the landing products are similar with minor differences.
Robot boom, capital investment, company creation, a lot of products landed at auction, the industry seems to be bustling, but it is a mixed bag. The price of an intelligent security robot on the market is about 200,000, but you can search for similar products that are much lower than this price.

In April 2014, Huawei’s Bantian Industrial Park introduced a patrol security robot developed by CIIC. This robot is equipped with a 360° pan-tilt camera, which can realize functions such as face recognition, license plate recognition, and sound analysis.Five years have passed, technology application, product landing, market-driven, understandable, but some unscrupulous manufacturers have crowded into the market indiscriminately, vying to imitate and plagiarize, vaguely setting off a price war, disrupting the market, which is obviously not conducive to the further development of the entire industry. In this homogeneous market environment, it is difficult to innovate and break through to achieve the desired state of security.

Of course, intelligent security robots are not in a slump, and there are still breakthroughs and improvements in some modules. UBTECH's ATRIS has achieved good development in chassis optimization. The robot adopts a crawler-type chassis, with independent shock absorbers hanging on the left and right, and can climb stairs and work in complex environments such as soft and muddy ground. However, this is just a progress that cannot reach the core for security robots. Monitoring upgrades and security operations are the core business of security robots.

Intelligent security robots are " better off than many ", and the future of security can be expected.

Intelligent security robot is the landing product of "AI + robot + security". AI is the brain, responsible for data processing and feedback; robot is the body, responsible for interaction and collaboration with the scene; security is the goal and the behavior orientation of "AI + robot". The combination of science is the result of 1+1>2, and intelligent security robots are expected to solve the pain points that currently plague the security industry. The security environment is complex. Taking the passenger flow at Shanghai People's Square subway station as an example, the official data can reach an average of 786,000 people per day. Even if the digitization of the physical world can be realized, there is still a huge amount of information of nearly 790,000 pieces of data, and it is almost difficult for humans to find key information in these data groups.

First, the security environment is complex. Taking the passenger flow at Shanghai People's Square subway station as an example, the official data can reach an average of 786,000 people per day. Even if the digitization of the physical world can be realized, there is still a huge amount of information of nearly 790,000 pieces of data, and it is almost difficult for humans to find key information in these data groups.

But AI can analyze and extract these huge and complex data, and feed back to humans with a more intuitive result. Xiaojiu Data, which appeared on CCTV's "Intelligent", successfully completed the challenge of predicting the number of diners in the cafeteria by analyzing the big data of nearly 20,000 teachers and students of Northwest University, demonstrating the amazing algorithmic capabilities of AI.

In traditional cognition, a complex environment lies in the flow of people, which cannot be controlled and identified by humans. However, AI in the era of big data is very good at digitizing the real environment, and then simplifying the data, allowing security personnel to effectively identify risk factors in complex environments, lock security targets, and improve security.

Second, the efficiency of security is low. Working hours can be eight hours, but security must be continuous around the clock. This means that the pressure of security not only comes from the complex environment, but also has a high duration. In this respect, machines are superior to humans.

UBTECH's ATRIS adopts a crawler-type platform, which can work on soft, muddy and other complex ground, and can also climb 30° stairs, with good off-road maneuverability; visible light + infrared imaging, suitable for monitoring in any weather; 8h battery life, 2 hours of charging time, no emotions, no troubles such as illness, the intelligent security robot is fully capable of continuous work for 24 hours.

Third, the cost of labor is increasing. With the development of technology and economy, no matter which industry, the cost of labor is increasing and the cost of machines is decreasing. When the costs of labor and machines are at a certain intermediate point, there will be a large-scale phenomenon of machines replacing labor. In the security industry, due to the long-term maintenance of working hours and higher labor costs, there is an urgent need for "AI + security" products to be launched to solve the problem of high costs in this industry from a cost perspective.

However, at present, the launch of the security robot has not solved this pain point. On the contrary, because it is in the embryonic stage of the industry, the high scientific research cost makes the product temporarily unable to be popularized. The price of an intelligent security robot with better performance on the market is more than 200,000, which is higher than the labor cost. Based on the average salary of the security industry in Changsha, the annual salary of one person is about 40,000 yuan, and 3 people work in 8-hour shifts around the clock, which means the annual cost is 120,000 yuan. Such an estimate is relatively vague and excludes multiple influencing factors, but it can be used as a reference for the current problem of high robot prices, and it is enough for us to see the cost gap between machines and labor. With the advancement and maturity of technology, we can foresee that this gap will be narrowed to the desired degree of control.

It is still in the early embryonic stage of the entire robotics industry, and the benefits that intelligent security robots can bring are not enough to shake the entire industry model. Human-oriented is still the main direction of security.

Products need to be polished and grown. The performance of intelligent security robots in terms of efficiency, monitoring and data processing is gradually superior to that of humans, which has improved the security of security. Through cooperation and assistance with humans, the optimal solution for the security industry can be obtained.

Then, the landing and application of AI+ robots in security scenarios is a promising future.