Path finding and obstacle avoidance in ant colonies
Duration and stipend:6 months, 500-600 euros/month Starting date: February-March 2021
Internship institution: LISN, équipe ParSys, physically in LRI (Laboratoire de Recherche en Informa- tique), Université Paris-Saclay (remote supervision in case of COVID lockdown is possible)
Supervisors: Evangelos Bampas (bampas@lri.fr), Joffroy Beauquier (jb@lri.fr), Janna Burman (bur- man@lri.fr)
Keywords:distributed nature-inspired algorithms, fault tolerance
To apply:contact the supervisors presenting your motivations and relevant skills (preferably including grade transcripts)
Many insects depend on pheromones to communicate with each other, whether searching for food or finding a mate. Insects can detect existing pheromone in their immediate neighborhood by “smelling”
with their antennae. Different types of ants also possess limited long-range sensing capabilities, including long-range vision and extraction of directional information out of the vibrational signal of stridulation.
Such biological observations lead to incorporate in an ant colony model a long-range detection capacity.
We naturally assume that long-range detection consumes more resources than short-range sensing, so that long-range detection must be used as little as possible.
The goal of this internship is to obtain distributed algorithms using both short-range and long-range detection, which would allow to better understand how nature works, or at least to produce executions that would be similar to those observed in the nature. These algorithms or their underlying ideas could influence solutions of distributed problems in the most efficient way, including time efficiency and resource consumption.
The approach will be formal and not experimental, although simulations could be used for refining or finding the best solutions. It consists in defining a model, proposing algorithms in this model, proving their correctness, and analyzing their performance. After having specified an adequate model, the next step will be to solve a particular problem in this model. An informal description of the problem follows.
There is a treasure at some position of a two dimensional space. Initially, all ants but one are in the nest. The single ant is at the treasure. One can imagine that it has found the treasure by a random walk, or by using already known treasure hunting algorithms. The role of the single ant is to make possible long-range detection of the treasure. The first goal is to establish a unique pheromoned path (a path along which all positions have a strictly positive amount of pheromone) between nest and tresor, which must be stable, despite pheromone evaporation. Thus the path has to be constantly reinforced by the ants walking on it, once it has been built. The second goal is related to failure tolerance. Obstacles can appear, but their number is finite. Once an obstacle appears, it stays still and unmodified for the rest of the execution. An obstacle may appear before or after a path is built. Then, if a path is interrupted by the arrival of an obstacle, another path has to be built. We assume that the size, the shape and the positon of an obstacle are chosen by an adversary. Nevertheless the system has to eventually converge towards a stable pheromoned path avoiding all obstacles.
Related problems have been studied in the literature both from a biological perspective ([1, 2]) and an algorithmic perspective ([3, 4]). Existing work in the distributed algorithms literature considers ant communication via pheromones, but pheromones are viewed as static non-evaporating markers left by the ants on graph nodes. We aim to produce the first formal algorithmic results that take into account pheromone evaporation.
References
[1] A. Chandrasekhar, D. M. Gordon, S. Navlakha: A distributed algorithm to maintain and repair the trail networks of arboreal ants. Scientific Reports 8, 9297 (2018).
[2] E. J. T. Middleton, T. Latty: Resilience in social insect infrastructure systems. Journal of the Royal Society Interface 13: 20151022 (2016).
[3] O. Feinerman, A. Korman: The ANTS problem. Distributed Computing 30, 149-168 (2017).
[4] C. Lenzen, T. Radeva: The power of pheromones in ant foraging. 1st Workshop on Biological Distributed Algorithms–BDA 2013.
