Autonomous Security Robots

Autonomous security robots have the advantage of performing their tasks without human intervention, thus alleviating the burden on humans by repetitive chores or lengthy operations hours. An excellent example of such robots are those used in security, as these robots do not require sleep and do not lose focus due to extended working hours. In the following subsections, we will examine robots without an omnidirectional camera as well as those employing an omnidirectional camera in their operation.

2.1.2.1 Autonomous Security Robots Utilizing a non-Omnidirectional Camera

Many security robots perform their operation without utilizing an omnidirectional camera. They depend on other sensors to carry out their tasks such as pan-tilt-zoom (PTZ) cameras, thermal cameras and other sensors.

In [21] and [22], multisensory fusion is used for environment monitoring, motion detection and surveillance using a mobile robot equipped with a number of sensors. In [23] a set of three robots, each equipped with a CCD camera, ultrasonic sensor, fire-detecting module and an RFID based localization unit, are used to detect fire, gas and radiation. When detected, the robot sends a message to a web-control center and another robot comes to confirm the accident.

In [24] a sensor network is used to detect intruders. When an intruder is detected, the robot is instructed to jump towards the sensor location and start taking photos once it gets there. The photos are sent to a gateway which in turn sends them to the user.

In many systems, a simple autonomous robot is used to monitor an environment and report any abnormal events such as a sound, motion, fire or other. Examples of such robots are exhibited in [25][26][27][28][29][30][31][32] .In [25] and [26] a robot uses a camera as well as a set of microphones to detect abnormal activity such as fighting, running or other behaviors that may produce certain sounds inside a home, verify the source of the sound and report an image to the master’s mobile phone. In [27] a robot uses a microphone as well as a networked camera to detect intruders and report to a local monitoring station. It can also report to a remote monitoring station using the internet. In [28], [29], [30] and [31] the robot is also used to monitor an indoor location using a camera or other sensor and when an intruder or other event is detected the user is notified through an SMS or email, and in [32] a dog-shaped robot was tested in a supermarket environment to interact and report accidents and events to its supervisor center.

In [33] two robots, one fully autonomous and one partially autonomous were used to patrol an indoor environment. The autonomous robot is equipped with an IR sensor for detecting intruders and RFID scanner to verify them. If the intruder fails to present a correct RFID tag, then the robot will attack the intruder, otherwise it will continue its normal operation.

In [34] a robot operates autonomously to map and then navigate the environment. The robot utilizes three-layer control architecture for primitive behaviors, complex tasks and control algorithms. The robot is programmed to detect missing object and detect intruders, when an intruder is detected, the robot will follow him/her until the intruder stops or disappears from the scene.

In [35] and [36] a robot with a thermal sensor is used to detect the presence of a person, and another camera is used for the face detection and recognition purposes and in [37] a robot navigates a pre-determined path and uses a camera to recognize the face of an intruder and an alarm is raised if the face is not recognized.

In [38] an autonomous system incorporating a camera-on-rail is used to monitor an indoor environment and detect intruders. The merit of such a system is that it does not require any obstacle detection or complex navigation capabilities as

well as being non-obstructive to humans. The disadvantage however, is the inability to track a subject beyond the limits of the rails being used.

In [39] a mobile platform is controlled using a number of stationary cameras. The direction of motion is decided based on the location of the person to be tracked or followed and the presence of an obstacle which is detected by the robot. The problem with this system is that it will be incapable of tracking a subject beyond the scope of the stationary security cameras.

Although it is possible to design and use a robot without employing an omnidirectional camera, it is obvious that such a sensor would increase the robot’s awareness of its surrounding and possibly reduce the number of sensors onboard or exterior to the robot, as well as improve the robot’s ability to track a subject and apply suitable navigation and obstacle avoidance methods.

2.1.2.2 Autonomous Security Robots Utilizing an Omnidirectional Camera

Robots that utilize an omnidirectional camera are able to detect motion or environment variations in any direction without having to turn or maneuver. In [40] five different types of sensors were used; a video motion detector, a passive infrared array, an acoustic sensor array and an ultrasonic array. Each sensor arrays covered 360 degrees and was independent of other sensor arrays. The information from the sensors was fused to reach a conclusion of an intruder detection, which would result in raising an alarm. Although the authors report a 99% success rate in detecting intruders, the presented design falls short of recognizing the person’s identity. An intruder can be one of the security personnel happening to pass by for any reason.

In [41] an autonomous sea surface vehicle (ASV) uses a set of 6 cameras to capture a 360 degree view to monitors an area searching for other vessels and determine whether it’s adversarial. Such a sea robot (in the form of an autonomous ship) is highly preferential as it relieves service personnel from enduring different conditions at sea as well as being on guard 24/7. The efficiency of such a platform, however, is highly dependent on the sensors provided and the algorithms used.

In [42] a robot is trained manually to follow a certain route, during which it registers the environment by capturing a panoramic image every few centimeters. Later, the robot follows the same path and compares the environment images captured by the camera with those previously stored, if a difference (anomaly) is detected, then an alarm is raised to alert a security person to intervene. This approach would be robust in situations where the environment does not change such as a night shift at a museum, library or office, but would be unsuitable for situations where the environment is volatile due to the addition or removal of any item within, or even the passing by of a security worker for any reason.

2.1.2.3 Comparison of Autonomous Security Robots

To better highlight possible research gaps, Table 2.1 provides a general comparison between the robot presented in this work (highlighted in blue) and previous robots mentioned in the literature.

Heading Description

Ref. No. : The reference number in which this robot was mentioned

Robot Name : The name of the robot; if any.

Area of Operation : Indoors, outdoors, underwater, factory … etc.

Vision Coverage : The area covered by the robot’s vision system: e.g.: 90,

180, or 360 degrees.

Size : Small (5 - 35), Medium (36 - 75), Large (76 - 175)

Huge (176 - 300), Gigantic (over 300) cm

Speed of Operation : Fast, slow, or the actual speed if mentioned in the literature.

Sound Det. : Sound detection capability.

Motion Det. : Motion detection capability.

Face Det. : Face detection capability.

Face Rec. : Face recognition capability.

Authent. : Intruder authentication / verification capability.

Sensors Used : The sensors used by the robot

Table 2.1: Comparing the robot in this work with robots mentioned in the literature

Legend

* Not mentioned, therefore anticipated based on the robot design

** Pan & Tilt, not a built in viewing angle

*** Omnidirectional camera

Ref.

No. Reference Title Robot Name Area of Operation Visoion Coverage (Deg) Size Speed So un d D et . M ot io n D et . Fa ce D et . Fa ce R ec . A ut he nt .

Sesors Used Problem / Research Gap

-

Face Detection Based Autonomous Security Robot Incorporating Omnidirectional Vision