Distributed, deterministic control for restoring and improving k-connectivity in multi-robot networks — the implementation accompanying our IROS 2026 paper.
Atharva Sagale · Tohid Kargar Tasooji · Ramviyas Parasuraman HeRoLab, University of Georgia Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2026
Multi-robot coordination requires maintaining network connectivity, especially in critical operations such as search and rescue, where network robustness is paramount. In this paper, we study the Fast k-connectivity Restoration problem (FCR), which aims to minimize the maximum distance required to restore connectivity. Recent works have proposed scalable solutions, but they rely on a centralized architecture or a learning-based solution to transfer the observable policies from centralized to distributed variants. However, computing node connectivity in a distributed setting is challenging, and performing such operations in a deterministic, algorithmic manner is critical for generalizability and persistent deployment in diverse, unknown environments. We propose a distributed control approach for improving and restoring connectivity. Using local neighborhood interactions and information, our algorithm improves the degrees of individual robots by augmenting 1-hop edges based on their distances, achieving the desired level of node connectivity and yielding comparable performance to achieve sparse connectivity improvements and up to 58% reduction in movements required to achieve dense connectivity over the state-of-the-art algorithms.
Figure 1. Depiction of the FCR problem showing the initial configuration and the final configuration after augmented edges (links) to restore k-connectivity to K = 4 or 6.
Figure 2. Working of the proposed approaches and baselines. The figures illustrate the final positions of the robots after edge augmentation and displacement. The edges in red represent the augmented edges.
Figure 3. Real-world experiment showcasing DEA-DR-k on our in-house swarm robotics testbed with a team of N = 8 robots, sensing radius = 0.5 m.
| File | Description |
|---|---|
DEA-DR-k.py |
k variant (DEA-DR-k): augments and verifies node connectivity κ ≥ K; also exposes the single-pass DEADR. |
DEA-DR-d.py |
δ variant: single-pass degree augmentation (no connectivity check). |
figures/ |
Paper figures. |
Requires Python 3 and the following packages:
pip install numpy matplotlib networkx pandas# k variant (DEA-DR-k): iterates until node connectivity reaches K
from DEA_DR_k import DEA_DR_k
out = DEA_DR_k(seed=35, N=10, delta=0.5, K=4, plots=True)
# delta variant: single-pass degree augmentation
from DEA_DR_d import DEA_DR_d
out = DEA_DR_d(seed=35, N=10, delta=0.5, K=6, plots=True)Parameters
| Parameter | Type | Description |
|---|---|---|
seed |
int |
RNG seed for the random deployment |
N |
int |
Number of robots |
delta |
float |
Sensing radius δ |
K |
int |
Target degree / connectivity |
plots |
bool |
Show the figures |
locations |
ndarray |
Optional 2 × N array overriding the random deployment |
Returns The augmented positions, max / total robot displacement, the initial vs. final node connectivity, edge connectivity, and minimum degree, plus the number of edges added and relocation steps (see each function's docstring for the exact tuple order).
Running either file directly (python DEA-DR-k.py) shows the plots and holds the
windows open until a key press; when the functions are called from a notebook/REPL
cell, figures render inline and there is no wait.
- Atharva Sagale - Graduate Student
- Dr. Ramviyas Parasuraman - Lab Director
School of Computing, University of Georgia.
For further information, please contact Dr. Ramviyas Parasuraman at ramviyas@uga.edu.