Robots on Patrol, to Guard and Protect

January 2015
Original Source: MIT Technology Review

Image Source: Knightscope

Microsoft has hired five units of the autonomous security robot K5, manufactured by Knightscope, to patrol its campus in Silicon Valley. The robots’ mission is to predict and prevent criminal activity.


  • The First Autonomous Data Machine – This is how Knightscope defines the K5, which features a 360° high definition camera, radar, lidar, thermal imaging, sensors for smells, speed, distance and air quality, along with letter-scanning technology capable of reading and recording up to 300 vehicle license plates per minute.

  • Patrol, Detect and Warn - K5 moves about autonomously within a given perimeter by randomly chosen routes so that its movements are not predictable. It harvests data through its sensors, and combines them with other information sourced from businesses, government and social networks by sophisticated algorithms, all in order to calculate the likelihood of criminal activity, detect dangerous situations and issue alarms in real time.

  • Machine as a Service - The business model for the K5 is the Machine as a Service one, with a rental cost of $3,000 per month for 24/7/365 service. Knightscope also foresees bringing in one, two and three year subscription plans for interested companies.


  • Multipurpose Robots whose Arsenal is Sensors not Weapons  - Moving at a speed of 8 kilometres per hour over flat surfaces, the K5 is intended to be a deterrent rather than a means of reacting to possible criminal activities. The K5 carries no weapons but is equipped with a full array of sensors which make it a self-contained detection system capable of seeing, hearing, smelling, recognizing and remembering unauthorized intrusions. Depending on the combination of sensors selected by the user, the robot can be endowed with different capabilities suited to different needs in different situations, depending on whether it is deployed in public spaces, schools, warehouses, offices, airports or other places. In future, wherever a need for vigilance or monitoring by land, sea or air exists, there will be robots to carry it out. The technology of the sensors and the data-recording is evolving but is already mature enough for us to predict with confidence that within the next five years robotic patrols will cease to be objects of curiosity and will become fully integrated into our human surroundings.

  • The Demand for Security Personnel Continues to Increase - According to estimates from the US Department of Labor's Bureau of Statistics there were more than a million security guards employed last year in the United States. It is a sector with a high turnover, where the average cost per man-hour is $13.24, double the cost of a K5, which is estimated at $6.25 per hour. However, for now, no security guard can do what the K5 does with the same efficiency, and no robot can do what a guard does with the same versatility. The functions of both robot and human are complementary, and it is too early to speak of a widespread loss of jobs with the introduction of robot patrols, at least in the medium term. On the contrary, according to ASIS International, heightened threat levels for theft, sabotage and other dangers have stimulated increased demand for workers who are qualified in security management.

  • The Privacy of Personal Data - Unauthorized data capture and the invasion of personal privacy are hot topics of interest to the general public which generate enormous controversy. In order to detect possible threats and raise the alarm, if necessary, the K5 obtains data from its own multiple sensors, but also from businesses, governments and social networks, in real time. Knightscope argues that its robots are mostly deployed in public spaces and that the K5 is nothing more than a more visible alternative to the ubiquitous security cameras. But there are still many unknowns to do with the treatment of data by private companies which will have to be addressed. While surveillance cameras are legal, the recording of speech, for example, is not regulated. What guarantees as to the security of private data does the K5 offer? What happens to its data if the device is stolen? Who will be able to access our discussions and conversations, and for what purposes? As long as there are no clear regulations protecting the privacy of personal data collected by terrestrial and aerial robots, some will recall the quote attributed to Benjamin Franklin: "Those who surrender freedom for security will not have, nor do they deserve, either one."

  • Robots which Learn in Changing Environments - The majority of autonomous robots are equipped with maps of their surroundings or are able to create them for themselves at the start of their activity. In environments which are constantly changing, such as are most human environments, the robot must update its map periodically, which lessens its operative autonomy and functionality. The Linda robot from the University of Lincoln in the UK has been designed to overcome the challenges of changing environments by learning its surroundings in a methodical way. Through continuous observation of people, shapes and activities, Linda "learns" to recognize the constants in any given area and is thus able to identify unusual activity within it and to act accordingly. Linda is one of seven robots created under the EU STRANDS program, which is aimed at creating autonomous intelligent robots capable of operating for months at a time in changing human environments.

  • What Will the Security Guards of the Future be Like? - Linda sets the trend for how and what the surveillance and monitoring robots of the future will be. We'll see robots with a long battery life (Linda can now run for 30 days in a row) which, through pattern recognition and continuous learning from their environment, will be able to spot suspicious changes in routines or unusual behaviors (an object changing position, an open door, an intruder in a restricted area, an old person collapsed in a hallway ...). So it will be able to monitor and collect data from its environment while at the same time recognizing, for example, that a door which is open should be closed, thus acting in much the same way that a human would. For now though, Linda is still too expensive for general use, at more than $40,000 per unit. Its deployment remains restricted so far to relatively predictable environments, such as museums, nursing homes or collaborations with human security guards. Its roll-out onto a wider market is likely to take place over the next decade as its price comes down and its functionality improves.