Data from: Self-organizing nervous systems for robot swarms

Zhu, Weixu1 ; Oguz, Sinan1 ; Heinrich, Mary Katherine 1 ; Allwright, Michael1 ; Wahby, Mostafa1 ; Lyhne Christensen, Anders2 ; Garone, Emanuele1 ; Dorigo, Marco 1

Published Nov 05, 2024 on Dryad. https://doi.org/10.5061/dryad.xpnvx0knc

Abstract

We present the self-organizing nervous system (SoNS), a robot swarm architecture based on self-organized hierarchy. The SoNS approach enables robots to autonomously establish, maintain, and reconfigure dynamic multilevel system architectures. For example, a robot swarm consisting of n independent robots could transform into a single n–robot SoNS and then into several independent smaller SoNSs, where each SoNS uses a temporary and dynamic hierarchy. Leveraging the SoNS approach, we showed that sensing, actuation, and decision making can be coordinated in a locally centralized way without sacrificing the benefits of scalability, flexibility, and fault tolerance, for which swarm robotics is usually studied. In several proof-of-concept robot missions—including binary decision making and search and rescue—we demonstrated that the capabilities of the SoNS approach greatly advance the state of the art in swarm robotics. The missions were conducted with a real heterogeneous aerial-ground robot swarm, using a custom-developed quadrotor platform. We also demonstrated the scalability of the SoNS approach in swarms of up to 250 robots in a physics-based simulator and demonstrated several types of system fault tolerance in simulation and reality.

https://doi.org/10.5061/dryad.xpnvx0knc

Authors

Weixu Zhu (ULB, co-first author)
Sinan Oguz (ULB, co-first author)
Mary Katherine Heinrich (ULB, co-first author)
Michael Allwright (ULB)
Mostafa Wahby (ULB)
Anders Lyhne Christensen (SDU)
Emanuele Garone (ULB)
Marco Dorigo (ULB)

*Contact (Corresponding authors)

mary.katherine.heinrich@ulb.be
mdorigo@ulb.ac.be

This dataset accompanies the article Self-organizing nervous systems for robot swarms: https://doi.org/10.1126/scirobotics.adl5161

Description of the data

The dataset is organized according to experiment type and contains folders of results from real robot experiments and experiments in simulation:

Each folder containing real robot experiments contains data from 5 or 6 trials.
Each folder containing simulation experiments contains folders for different system sizes (i.e., number of robots), which each contain data from 30 or 50 trials.
Some folders contain demos

For details of the experiment types and setups, please refer to the accompanying article on self-organizing nervous systems (SoNSs): https://doi.org/10.1126/scirobotics.adl5161

File structure

The directories containing results data are organized as follows:

1_Mission_Self-organized_hierarchy

Variant1_Clustered_start
- Real_robots_experiments
- Simulation_experiments
  - 8robots
  - 50robots
Variant2_Scattered_start
- Real_robots_experiments
- Simulation_experiments
  - 12robots
  - 50robots

2_Mission_Global_local_goals

Variant1_Smaller_denser_obstacles
- Real_robots_experiments
- Simulation_experiments
  - 8robots
  - 50robots
Variant2_Larger_less_dense_obstacles
- Real_robots_experiments
- Simulation_experiments
  - 8robots
  - 50robots

3_Mission_Collective_sensing_actuation

Real_robots_experiments
Simulation_experiments
- 8robots
- 50robots

4_Mission_Binary_decision

Real_robots_experiments
Simulation_experiments
- 8robots
- 65robots

5_Mission_Splitting_merging

Variant1_Search_and_rescue_in_passage
- Real_robots_experiments
- Simulation_experiments
  - 8robots
Variant2_Push_away_obstruction
- Real_robots_experiments
- Simulation_experiments
  - 5robots
Large_scale_Simulation_experiments
- 50robots

6_Scalability_setups_a

Scalability_in_decision_making_mission
35robots
65robots
95robots
125robots

6_Scalability_setups_[b-h]

Scalability_in_SoNS_establishment_mission
[5-250]robots

7_Fault_tolerance_setups

Real_robots_experiments
Simulation_experiments
- 8robots
- 50robots
  - Variant1_One_third_chance_for_each_robot_to_fail
  - Variant2_Two_thirds_chance_for_each_robot_to_fail
  - Variant3_Temporary_system_wide_vision_failure
    - 0.5s_failure
    - 1s_failure
    - 30s_failure
  - Variant4_Temporary_system-wide_communication_failure
    - 0.5s_failure
    - 1s_failure
    - 30s_failure
  - Variants_additional
    - One_third_chance_for_each_aerial_robot_to_fail
    - One_third_chance_for_each_ground_robot_to_fail
    - Two_thirds_chance_for_each_aerial_robot_to_fail
    - Two_thirds_chance_for_each_ground_robot_to_fail

8_Demos

Aerial_robot_hover_in_simulation
Ground_robot_brain
- 3_Mission_Collective_sensing_actuation
Random_quality_values_at_initialization
- 1_Mission_Self-organized_hierarchy
- 3_Mission_Collective_sensing_actuation
Real_robots_with_failure_and_substitution
- Brain_aerial_robot_failure
- Ground_robot_failure
- Non_brain_aerial_robot_failure

Each of the above directories contain one folder for each trial, labeled as: run[1-n]

Each run[1-n] folder contains the following:

NOTE: in all csv files, each column corresponds to a variable and each row corresponds to a time step (in ascending order, starting with time step 0).

A. experiment_data folder

folder includes one csv file for each robot included in the experiment trial, labeled using the ID number of the robot, as:
- drone[ID].csv if the robot is an S-drone quadcopter
- pipuck[ID].csv if the robot is an e-puck mobile robot with a pi-puck extension board
variables in the drone[ID].csv and pipuck[ID].csv files:
- position_x, position_y, position_z: x,y,z position of the robot recorded by the external motion capture system (meters)
- orientation_x, orientation_y, orientation_z: x,y,z orientation angles of the robot recorded by the external motion capture system (Euler angles)
- virtual_frame_x, virtual_frame_y, virtual_frame_z: robot’s self-recorded x,y,z orientation angles of its intermediary motion frame relative to its body frame (Euler angles) (for an explanation of the intermediary motion frame, please refer to the Supplementary Materials of the accompanying article)
- goal_position_x, goal_position_y, goal_position_z: robot’s self-recorded x,y,z target position relative to its intermediary motion frame (meters)
- goal_orientation_x, goal_orientation_y, goal_orientation_z: robot’s self-recorded x,y,z target orientation angles relative to its intermediary motion frame (Euler angles)
- target_id: robot’s self-recorded node ID in its SoNS target graph
- brain_name: robot’s self-recorded ID of the SoNS that it belongs to (i.e., the robot ID of the brain of the SoNS it belongs to)
- [OPTIONAL – column only included for some simulated experiments] parent_name: robot’s self-recorded ID of its parent, if it has one (i.e., the robot ID of its parent, if it has one)

B. error_measurements folder

error_per_robot folder
- includes one csv file for each robot included in the experiment trial, labeled using the ID number of the robot, as above
- each drone[ID].csv or pipuck[ID].csv contains one column giving: the actuation error of the robot (please refer to Eq. 1 in the accompanying manuscript)
error.csv, which contains one column giving: the mean actuation error of all robots in the experiment (please refer to Eq. 2 in the accompanying manuscript)

C. _failed_robots.txt [OPTIONAL – only included for some experiments]

contains a list of the robot IDs of all robots that were subject to failure in that trial

D. _timesteps_target_SoNS_change.txt [OPTIONAL – only included for some experiments]

contains a list of time steps at which a brain triggered a change in the target graph of its SoNS

The remaining directories contain videos of the experiments/demos and plots of the results data in the directories above, organized as follows:

9_Videos

1_Mission_Self-organized_hierarchy
- Variant1_Clustered_start
- Variant2_Larger_less_dense_obstacles
2_Mission_Global_local_goals
- Variant1_Smaller_denser_obstacles
- Variant2_Larger_less_dense_obstacles
3_Mission_Collective_sensing_actuation
4_Mission_Binary_decision
5_Mission_Splitting_merging
- Variant1_Search_and_rescue_in_passage
- Variant2_Push_away_obstruction
6_Scalability_setups
7_Fault_tolerance_setups
- Demo_replace_a_robot
8_Demos
- Ground_robot_brain
  - 3_Mission_Collective_sensing_actuation
  - 6_Mission_Scalability_in_SoNS_establishment

10_Plots

1_Mission_Self-organized_hierarchy
2_Mission_Global_local_goals
3_Mission_Collective_sensing_actuation
4_Mission_Binary_decision
5_Mission_Splitting_merging
6_Scalability_setups
7_Fault_tolerance_setups

Code/Software

The code to run the experiments is primarily in C++ and Lua. The code to analyze the results data is primarily in Python. The simulation experiments require the ARGoS simulator and the real robot experiments require ARGoS libraries to be installed on the robots. The code and accompanying information is provided in the Zenodo software repository.

Supplemental information

The following documentation is provided in the Zenodo supplemental files repository:

SoNS theoretical analysis and mathematical proofs
Real indoor arena setup