Simulation, robot codes and figure data from collective phototactic robotectonics
Abstract
The collective construction of complex architectures by social insects via stigmergy is known to be modulated by spatio-time signals that modulate and are modulated by the environment. Inspired by these observations, we show that a robot collective can successfully nucleate a construction site via a trapping instability and then cooperatively either construct organized structures or de-construct them by (modifying a single parameter associated with the pick-up action of the robot). We quantify these observations in terms of a two-dimensional phase space, encompassing agent-agent interaction (cooperation) and the agent-environment interaction (collection and deposition). Our approach to complex task execution eliminates global representation, planning or optimization/control algorithms in favor of local rules for sensing and action. We observe that collectives can leverage the environment as both a communication channel and a spatio-temporal memory, which is likely to be applicable to broader contexts associated with embodied intelligence(s).
https://doi.org/10.5061/dryad.05qfttfb2
This dataset contains A) simulation codes, B) programs encoded in the robots, C) figure data and the calculation criteria for trapping of single and multiple agents in the article Giardina, Prasath et al. (https://doi.org/10.48550/arXiv.2208.12373.
Description of the data and file structure
There are three folders in the dataset: a) Figs, b) Code, c) Self-Trapping.
a) Figs Folder: The Figs folder contains all the data used to generate the plots in the main-article. There is a sub-folder titled Data which has the raw data of the plots in .txt/.csv format and is used inside the Figs.ipynb, a python notebook with detailed description as well as the plots used in the article.
b) Code folder: This folder has 5 sub-folders each with Arduino IDE code (inside base/) for each robot to perform 1) Construction, 2) de-construction, 3) no gradient descent (dynamics without photormone following behavior), 4) no threshold (dynamics when there is no threshold for deposition of substrate elements), 5) pseudocode containing the algorithm of the code (also detailed in the main-article).
c) Self-trapping: Inside this folder there are 2 sub-folders: Calculation, Dynamic-trapping. 1) Calculation folder contains the Mathematica notebooks used to analyze the different asymptotic limits of self-trapping (detailed in the main-article as well as the supplementary materials). 2) Dynamic-trapping folder contains the Matlab code to simulation an agent following its photormone trail, the dynamics of which is solved using a finite-difference technique.