Supervisors info:
Ευστάθιος Χατζηευθυμιάδης, Καθηγητής, Πληροφορικής και Τηλεπικοινωνιών, ΕΚΠΑ
Summary:
In recent years, the rapid development and evolution of the Internet of Things (IoT) and robotics seems unstoppable. The new possibilities added to the nodes, open horizons for new research as well as for new uses in the daily life of people and industry. One of the key features is the mobility of the nodes. Most nodes are no longer static, but move in space and offer a wide range of new applications that can offer like the ability to make decisions without human intervention, their durability, the use of embedded sensors (temperature, pressure, humidity, etc.), as well as for their reprogrammability. Based on these features, mobile nodes can be used for example in cases of surveillance of areas and borders, for image recognition and alarm signaling, as well as for crisis management. For example, an unmanned land vehicle (mobile node) carrying a high-definition sonar and a high-definition thermal camera, combined with an object recognition algorithm can be used to find people trapped in wreckage.
In addition, this functionality can be enriched by the fact that two or more nodes can communicate with each other to work together to complete a mission. Let us consider a mission to find a lost hiker in a forest, with a single mobile node (unmanned aerial vehicle), the chances of finding him in a short time are much lower than when we have more than one to communicate with, exchanging images, measurements, the areas they have scanned, and finally if any of them have found the target. This group mode in the context of the Internet of Things and robotics is called a swarmi of nodes. More specifically, each node operates based on the knowledge of the whole team and not individually. This is also observed in nature, especially in insects, where they function on the basis of this method.
In this thesis, it is examined whether the swarm operation is more efficient both temporally and qualitatively in relation to the operation of each node as independent on collaborative search of sensor targets with no prior knowledge of the environment. More specifically, a series of experiments are carried out where two robots scan the entire space in detail exhaustively in order to identify the points where the value from an existing sensor sources is maximum. These values are detected by robots with the help of sensors that they carry. In the first case, the robots act independently without knowing neither the measurements taken by the other, nor its position. In the second case, the robots cooperate based on the operation of the swarm, in order to find the optimal possible value of the source.
The experiments were supported by the Ubuntu 16.04 operating system, the Gazebo and Rviz simulators, as well as two virtual TurtleBots running the ROS operating system, as well as a virtual XBOX Kinect sensor with a color camera and a depth sensor.
Keywords:
collaborative, context exploration, path planning, ROS, turtlebot, Particle Swarm Optimization
