Home

3D Multi-Robot Patrolling with a Two-Level Coordination 
Strategy


Teams of UGVs patrolling harsh and complex 3D environments can experience interference and spatial conflicts with one another. Neglecting the occurrence of these events crucially hinders both soundness and reliability of a patrolling process. This work presents a distributed multi-robot patrolling technique, which uses a two-level coordination strategy to minimize and explicitly manage the occurrence of conflicts and interference. The first level guides the agents to single out exclusive target nodes on a topological map. This selection relies on a shared idleness representation and a coordination mechanism preventing topological conflicts. The second level hosts coordination strategies based on a metric representation of space and is supported by a 3D SLAM system. Here, each robot path planner negotiates spatial conflicts by applying a multi-robot traversability function. Continuous interactions between these two levels ensure coordination and conflicts resolution. Both simulations and real-world experiments are presented to validate the performances of the proposed patrolling strategy in 3D environments. Results show this is a promising solution for managing spatial conflicts and preventing deadlocks.

  
                   

    


Documents 

Luigi Freda · Mario Gianni · Fiora Pirri
Abel Gawel · Renaud Dubé · Roland Siegwart · Cesar Cadena
"3D Multi-Robot Patrolling with a Two-Level Coordination Strategy"   
submitted to Autonomous Robots Journal



Simulations 

The following simulations are realized within V-REP. The environment is represented by using a point cloud map. The built map is segmented in traversable regions (green) and obstacle regions (red).


Path planning with multi-robot traversability 

The following two simulations show how our path planner succeeds in managing metric conflicts by using the multi-robot traversability. Each robot moves back and forth between its two assigned waypoints. Planned paths (emanated from each robot) are shown in the videos.
  • path planning and multi-robot traversability in a free area with 3 robots - video
  • path planning on cyclic paths with three robots
    • without metric coordination - video    (here the basic path planner does not succeed without metric coordination in avoiding a final deadlock)
    • with metric coordination - video    (metric coordination nicely allows the robots to avoid congestions and deadlocks)

Patrolling 

The following simulations show our full patrolling strategy with a team of robots. Here, the CC (Coordination Plus Cooperation) strategy is used in order to guide the robot team. Each node of the patrolling graph is represented by a disc/balloon whose radius grows with time proportionally to the node idleness. When a robot reaches a node, the local "balloon" is deflated (i.e. the idleness is reset). In short, the main mission of the patrolling robots is to plan their motions so as to continuously visit the nodes and keep the balloons as small as possible.
  • 3 robots in the small-crossroad scenario - video 
  • 4 robots in the crossroad scenario - video 
  • 4 robots in the corridor scenario  - video 
  • 4 robots in the fork scenario - video 
  • 4 robots in the multi-floor ramp scenario - video 
  • 4 robots in the two-floor ring scenario - video 

In our patrolling framework, you can set the priority of each node in the patrolling graph. The priority of a node controls the inflating speed of the local "balloon": the higher the priority, the higher the inflating speed. The following simulations show how the priority can be set by using the system GUI, and the typical behavior of the robots when node priorities are changed.
  • setting the priority of a node - video
  • 4 robots in the crossroad scenario when some priorities are changed - video  
N.B.: in the videos, speed can be increased by using the media player interface


Experiments

In the following experiments, the environment is represented by using a point cloud map. The built map is segmented in traversable regions (green) and obstacle regions (red). Each node of the patrolling graph is represented by a disc whose radius is proportional to the node idleness. Planned paths are emanated from each robot.


The following patrolling experiments aim at showing the applicability and portability of the developed software in the real world. The CC (Coordination Plus Cooperation) strategy was tested on the TRADR robots.
  • exp1: 2 robots in DIAG corridor with CC strategy - video (2X speed)
  • exp2: 2 robots in DIAG corridor: this experiment highlights the dynamic mapping capability of our system - video (2X speed)
  • exp3: 2 robots in DIAG court with CC strategy - video (2X speed)
Experiments with 3 UGVs:
  • exp4: 3 robots in DIAG corridor (big graph) with CC strategy - video (2X speed)
  • exp5: 3 robots in DIAG ramp with CC strategy - video (2X speed)
N.B.
 video speed can be further increased on the media player interface
 in the above videos, the visualized data in RVIZ may sometimes appear delayed: this is due to the transfer and visualization of the point cloud maps on the central user interface; the use of such a GUI is an option and is not necessary to the patrolling system

     




TRADR EU Project

This work was conducted within the TRADR EU project (No. FP7-ICT-609763). You can find a presentation of the project on its website

TRADR Review Year 4on March 21-22 2018, we successfully demoed our patrolling system in a rescue scenario at an industrial site in Mestre, Italy, under the great hospitality of our Italian Firefighters, Vigili del Fuoco.

    


TRADR Review Year 3: on March 8th 2017, we successfully showed two robots patrolling in a rescue scenario during the third year review of the TRADR EU Project. The review was held in Montelibretti at the "Scuola di Formazione Operativa di Montelibretti" of Vigili del Fuoco (the Italian Firefighters).
 


Here you are some video shots of the demo:
  • two robots patrolling in the gallery - video
  • patrolling around a wrecked car - video
  • patrolling in the dark - video






Authors

Luigi Freda · Mario Gianni · Fiora Pirri   (Alcor Lab - University of Rome "La Sapienza")
Abel Gawel · Renaud Dubé · Roland Siegwart · Cesar Cadena  (ASL ETHZ)


Credits 

We would like to thank Indro Spinelli for his great help in performing patrolling experiments with the robots. 




website maintained by Luigi Freda
Comments