Otto-von-Guericke-Universität Magdeburg

 
 
 
 
 
 
 
 

tornow

 Distributed-Environment-Perception-by-a-Multi-Agent-Robot-System

Robots can assist humans during a mission e.g. searching for victims in case of a catastrophe within dangerous environments as well as exploring areas for transportation of materials through rough terrain. A multi-agent system can speed up this process by working as distributed sensor system. Therefore the task is divided into subtasks each performed by an individual robot within the team.

For exploration of an unknown area a robot requires a positioning system and a data distribution network besides the sensor array. As test platform the Eddie robot equipped with an active optical sensor system (e.g. Microsoft Kinect) as well as additional IR and ultrasonic sensors is used. These sensors are used for environmental perception and obstacle detection. A WLAN-based network serves for the data distribution among the robots. Next to standard moving abilities the robot can to turn on it's vertical axis while maintaining position.

Text when figure is not displayed.
Schemata for Human-Machine-Interaction and searching for victims in an unknown environment; a) Robots team consisting of n+1 robots (R0... Rn), b) Searching efficiently for injured persons within an environment by avoiding multiple exploration of the same area

By giving all robots in the team equal HMI and environment perception capabilities a redundant system can be achieved. Furthermore by sharing information e.g. of parts of the environment which have already been explored over a private robot network can help avoiding multiple exploration resulting in increased efficiency: R2 moves towards area 3 instead of exploring area 2 where R1 is working). Meanwhile the operator may gives and distributes orders to the robot team.

For exchanging information regarding the surrounding area of the robots especially the 2D/3D information needs to be compressed due to communication channel limitations. To form a combined data structure of the environmental information of all robots it needs to be arrange with respect to the position and orientation of the individual robot. During this process a data compression by removing unnecessary redundancy is possible. Building a 3D panorama is one option for data reduction.

Contact: Michael Tornow , Ayoub Al-Hamadi

Letzte Änderung: 08.11.2017 - Ansprechpartner: Dipl.-Ing. Arno Krüger
 
 
 
 
image_pose
Video: Head pose and orientation
 
 
 
 
v01_winglets
Video: Particle Tracking
 
 
 
 
Johanniskirche_1_300x240
Video: Multi-object tracking
 
 
 
 
Test_584
Video: Gestures and intention
 
 
 
 
Kinect-pose-ayoub1-logo
Video: Pose and Face detection using Kinect
 
 
 
 
kinect_pose
Video: HCI Face Attention
 
 
 
 
mimik-flow1
Video: Static and dynamic features
 
 
 
 
Motionblobs-gut
Video: Trisectional Multi-object tracking
 
 
 
 
s8
Video: Ephestia Parasitization