Multi-security robotic system

vishnumuthu Avatar

About the author


MenuBar

 

Multi-security robotic system:

In recent years, autonomous security mobile robots have been widely studied to be used for scene surveillance and to understand, interpret, and anticipate human activities. In this article, a group of mobile robots patrolling office or home environment to understand the dynamic surrounding scene. First, the robot has to differentiate between the moving (humans, pets, etc.) and non-moving objects, for this purpose both 2D and 3D maps should be used. Mapping a dynamic environment is a critical task to provide better security evaluation. So, each individual robot should be able to detect and identify different types of objects / humans, and this process can be done by using traditional feature extraction or an advanced deep neural network. Second, Cooperative perception allows covering wide area and robust detecting the changes in the environment and, at the same time detection by more than one robot must be matched and coordinated for better understanding the surrounding.

When a single robot recognizes a static object in the scene, that object should be reconstructed in 3D model using the depth data. Then, the pose, orientation and the principal axis of the object is estimated based on the shape and structure, either it is symmetric or not. The modeled object is then labeled and localized on the map persistently. When a single robot recognizes a human or pet, the location is directly mapped on the 2D dynamic map, the human pose is estimated, and finally this information can be used to recognize the human state (standing, walking, etc.). When multiple robots detect the same object or human, there can be a disambiguate in the situation, the robot has to triangulate and figure out, whether it is the same or different object based on the location. During patrolling with multiple robots, with the help of the detected 3D model and reconstructions from each robot, object or human are separately marked in a global 3D map space in a server computer. Thus, detection of same object or human in a single robot is corrected and misalignment to global map is reduced. When an object or human moves in the scene during the focus of multiple robots, the change in the detection of the scene will help to track the object or human, based on the position of the map.

The table below describe different conditions for classification, mapping and tracking to provide better security systems,

Setting the alert based on the time period can give the robot a better perception of the environment. Below is the list of content for long term processing to understand the event based learning,

1) Human detection: Detection of humans when the robot is on patrol,

  • Robot on move (Dynamic State)
  • Robot at stopped position (Static State)

2) Human motion: When people run / walk,

  • Front of moving robot
  • Front of a stopped robot

3) Humans interaction: Human behaviour analysis,

  • People talking with each other
  • Identifying individual people action in a group

4) Human touching robot: Human behaviour analysis,

  • Robot identifying other robot
  • Robot being touched by human

5) Object detection: Detection of objects when the robot is on patrol,

  • Robot on move (Dynamic State)
  • Robot at stopped position (Static State)

6) Object place shift: When the object in the map is removed when the robot is not viewing (Place an object in map and remove it from the environment),

  • Single robot to identify
  • Single robot to search nearby area
  • Multiple robot to search later

7) Human – Object place shift (known): Human moving object (eg:chair) – remapping objects,

  • Front of single robot
  • Front of multiple robots

8) Human – Object place shift (unknown): Human moving object (unidentified) – remapping objects,

  • Front of single robot
  • Front of multiple robots

9) Unattended Object: Adding unifentified object, when robot view is absent,

  • Single / Mulitiple robots to detect and classification / identify

10)Human Detection at night: detect humans at unauthorized period,

  • Perform facial recoginition

There are wide range of applications that includes security systems in airports, super-markets, industrialized environment, etc.

“The robotic security system is essential to give humanity a peaceful sleep!”


Leave a Reply

Your email address will not be published. Required fields are marked *