SwarmHaul is a biologically inspired multi-robot system for collective object transport, motivated by the efficiency of ant colonies. Ants are known to move objects significantly larger than themselves with remarkable coordination — often outperforming humans. In contrast, many current multi-robot systems depend on extensive prior information such as object shape, precise initial formations, or centralized planning, limiting their real-world applicability.
This repository presents decentralized control rules for swarm-based object manipulation that require minimal assumptions:
- Only one robot (a "seed") needs to know the direction in which the object is located.
- The seed robot attaches to the object and recruits only two other robots through virtual stigmergy, starting two growing branches around the object.
- This process continue iteratively, where each robot in a branch recruits one another robot sequentially, eventually this process leads to caging the object completely.
- Once caged, robots cooperatively translate and rotate the object by maintaining a shared estimate of the object’s centroid — purely via local interactions (no external measurements or global coordinates required).
- ✅ Theoretical guarantees for caging any arbitrary convex objects
- 🤖 Decentralized coordination with no global position tracking
- 🔄 Simulations with up to 100 robots
- 🧪 Real-world experiments with 6 differential-drive robots
- 🧩 Extensible to heterogeneous swarms with onboard path planning (e.g., drones, AMRs)
For installation, you can either install everything from source or use the Docker installation
Follow the instructions in this page to install/build the programming language Buzz and Argos3 github.com/buzz-lang/Buzz/blob/master/doc/argos-integration.md
Install the khepera robot plugin for Argos3. These are the robots that will perform the collective transport https://github.com/ilpincy/argos3-kheperaiv
The order is important build Hooks_src before Loop_fun_src.
git clone http://git.mistlab.ca/vvaradharajan/collaborative_transport.git
cd collaborative_transport/src/Hooks_src/
mkdir build
cd build
cmake ../
make
cd collaborative_transport/src/Loop_fun_src/
mkdir build
cd build
cmake ../
make
You can build the docker in the docker docker folder. There is a build.sh that let's you do that. Finally you can run an interactive shell with sharing the display between the host computer for the gui with the run.sh
- Compile the buzz script:
cd scripts
bzzc -I includes/ Simulation.bzz
cd ..
- run the associated argos file
argos3 -c experiment.argos
# How to control the arguments in the experiment
<loop_functions library="src/Loop_fun_src/build/libplanning_exp.so"
label="Planning"
map_file_name="src/maps/Comparisions/empty.map"
map_option="0"
robots="ROBOTS" # No of robots check below
out_file="OUTFILE" # suffix to add to out files containg data for plotting
path="PATH" # can be straight, zigzac or straight_rot
inter_cage_dist="INTERCAGEDIST" # distance between caging
random_seed_set="RANDOMSEED" # random seed for the experiment
object_type="OBJECTTYPE"/> # type of object
Object type enum
0 - square object of size (2,2) for 25 robots
1 - square object of size (3.6,6) for 50 robots
2 - square object of size (7.2,18) for 100 robots
3 - cloud shape for caging tests
4 - box_rotation shape for caging tests
5 - clover shape for caging tests - needs 50 robots
- Improve readablity of
simulation.bzzfile - Make PyBuzz extension
- Support E-puck
- Support rolling spheres to be transported
This Ants video is taken from the paper:
"Ant groups optimally amplify the effect
of transiently informed individuals"
Aviram Gelblum, Itai Pinkoviezky, Ehud Fonio, Abhijit Ghosh, Nir Gov & Ofer Feinerman
Nature Communications, 2015
If you use this work, kindly consider citing us.
@InProceedings{10.1007/978-3-030-92790-5_27,
author="Vardharajan, Vivek Shankar
and Soma, Karthik
and Beltrame, Giovanni",
editor="Matsuno, Fumitoshi
and Azuma, Shun-ichi
and Yamamoto, Masahito",
title="Collective Transport via Sequential Caging",
booktitle="Distributed Autonomous Robotic Systems",
year="2022",
publisher="Springer International Publishing",
address="Cham",
pages="349--362",
isbn="978-3-030-92790-5"
}