On-line supporting material for "Communication assisted navigation in robotic swarms"

F. Ducatelle, G. Di Caro, C. Pinciroli, F. Mondada and L.M. Gambardella

 

 

Simple example in simulation
Dynamic chain in simulation
Shortest path in simulation
Real robots, test 1
Real robots, test 2
Real robots, test 3
Real robots, test 4

 
Abstract - We present a communication based navigation algorithm for robotic swarms. It allows robots to guide each other's navigation by exchanging messages containing navigation information through the wireless network formed among the swarm. We study the use of this algorithm in two different scenarios. In the first scenario, the swarm guides a single robot towards a target, while in the second, all robots of the swarm navigate back and forth between two different targets. In both cases, the algorithm provides efficient navigation, while being robust to failures of robots in the swarm. Moreover, we show that in the latter case, the system lets the swarm self-organize into a robust dynamic structure. This self-organization further improves navigation efficiency, and is able to find shortest paths in cluttered environments. We test our system both in simulation and on real robots.


Simple example in simulation

This simple example video, shot in simulation, explains how a swarm of randomly moving robots can guide a single robot to a target using communication assisted navigation. The target robot is in the bottom left corner (indicated with a green disk) and the searching robot starts in the top right corner (indicated with a yellow disk). Navigation information is spread through the swarm via network communication. It consists of a sequence number given out by the target, and a distance estimate. The searching robot moves towards the location of other robots which carry a higher sequence number or a lower distance. If a robot has useful navigation information for the searcher, this is indicated in the video with two bars next to the robot, a blue and a red one. The higher the blue bar, the better its sequence number. The lower the red bar, the shorter (better) its distance. The location the searcher is heading to is indicated with a red circle, while the path followed by the searcher is indicated by yellow dots. Whenever the searcher has no navigation information, it stops and waits.


Dynamic chain in simulation

This video gives an example of the dynamic chain behavior is simulation. Two targets are placed in the top right and bottom left corner (indicated by a blue and a yellow disk). 40 other robots are deployed randomly in the environment. Half of them go to one target, the other half to the other. When a robot reaches its target, it switches to the other target. The communication range of the robots is 3 m (the tiles on the floor have a side of 1 m). The swarm of robots collectively executing the communication based navigation behavior self-organizes into a stable and robust dynamic structure supporting efficient navigation.


Shortest path in simulation

This video gives an example of the dynamic chain behavior in simulation. Two targets are placed on opposite sides of the arena, with a large obstacle in between that causes two paths of different lengths to be available. The robotic swarm running communication based navigation self-organizes into a stable dynamic chain between the two targets. This chain forms preferentially over the shortest of two available paths, as in this example.


Real robots, test 1

This experiment shows how a swarm of randomly moving robots can guide a single robot between two target robots using communication assisted navigation. The searching robot uses the "navigation with stopping" strategy, meaning that whenever it has no information, it stops and waits.

We deploy a swarm of real foot-bots in a room of 12.7 x 3.4 m2. Two robots are placed on either end of the room to serve as targets. One robot moves back and forth between the targets while all other robots move randomly. The robots show their role with light signals, using the LED ring they have around their body, and the LED beacon they have on top. One target lights its ring and beacon in yellow, while the other lights them in blue. The searcher lights its ring in the color of the target it is moving to. It lights its beacon when it does not have navigation information. The randomly moving robots light their ring in red. We did the tests in the dark, so that the different colors are better visible. Unfortunately, sometimes robot LED rings stop working (e.g., after a colision), reducing robot visibility.

We start the experiment with 14 randomly moving robots, and gradually reduce this number by removing robots every 2 minutes. The experiment shows how the searching robot can move relatively smoothly, as long as the swarm size is large enough. When the swarm size becomes too small, the searching robot often needs to wait for new navigation information.

The video is shown with a speed four times faster than real time.

The tests were performed by F. Ducatelle and G. Di Caro. The foot-bot robots were developed at the EPFL LSRO lab by M. Bonani, S. Magnenat, P. Retornaz and F. Mondada.


Real robots, test 2

Here, we perform the same test as before, but using the "navigation with random" strategy, meaning that whenever it has no information, it moves according to the random direction mobility model. The experiment shows how with this strategy, the robot still navigates relatively smoothly when the swarm size gets quite low. When there are not enough other robots around, the searcher relies on random movement to find its target.

The tests were performed by F. Ducatelle and G. Di Caro. The foot-bot robots were developed at the EPFL LSRO lab by M. Bonani, S. Magnenat, P. Retornaz and F. Mondada.


Real robots, test 3

Here, we perform the same experiment as before, but in a different environment: one target remains at the end of the room, but the other is placed in the corridor adjacent to the room. The experiment shows how the size and complexity of the environment (moving through the door) form no problem for the system.

The tests were performed by F. Ducatelle and G. Di Caro. The foot-bot robots were developed at the EPFL LSRO lab by M. Bonani, S. Magnenat, P. Retornaz and F. Mondada.


Real robots, test 4

For this experiment, we place the target robots back at either end of the room of 12.7 x 3.4 m2. All other robots (15 in total) move back and forth between the targets, using communication assisted navigation.

The experiment shows how the swarm self-organizes into a dynamical structure, which allows efficient navigation between the targets. During the experiment, we first remove robots from the swarm, and see how the chain remains stable down to about 6 robots, below which robots move mainly randomly. Then, we place the robots back in the swarm and see the chain restore itself. We also move one of the target robots around twice, and observe how the chain finds the target back and restores itself to move in a straight line between the targets. At some point in the experiment, we remove a faulty foot-bot (its communication system had stopped working); this is only for esthetic reasons, the robot's faulty behavior did not disrupt the working of the system.

The video is shown with a speed four times faster than real time.

The tests were performed by F. Ducatelle and G. Di Caro. The foot-bot robots were developed at the EPFL LSRO lab by M. Bonani, S. Magnenat, P. Retornaz and F. Mondada.