arXiv++ Combinatorics

The ants process is a stochastic process introduced by Kious, Mailler and Schapira as a model for the phenomenon of ants finding shortest paths between their nest and a source of food (seen as two marked nodes in a finite graph), with no other means of communications besides the pheromones they lay behind them as they explore their environment. The ants process relies on a reinforcement learning mechanism. In this paper, we modify the ants process by having more than one ants nest (and still one source of food). For technical reasons, we restrict ourselves to the case when there are two nests, and when the graph is a triangle between the two nests and the source of food, whose edges have been replaced by series-parallel graphs. In this setting, using stochastic approximation techniques, comparison with Pólya urns, and combinatorial arguments, we are able to prove that the ants process converges and to describe its limit.

Consider a sequence of masses $m_0,m_1,...$ arriving uniformly at random at some points $u_0,u_1,...$ on the unit circle $\mathbb{R}/\mathbb{Z}$ (or on $\mathbb{Z}/n\mathbb{Z}$, in the discrete version). Upon arrival, each mass undergoes a relaxation phase during which it is dispersed, possibly also at random. This process can model many physical phenomena, such as the diffusion of liquid in a porous medium. In the discrete case, it can model parking (related to additive coalescence and hashing with linear probing) in which the cars are permitted to follow random displacement policies. The dispersion policies considered in the paper ensure that at time $k$, after the successive dispersions of $m_0,\cdots,m_{k-1}$, the total covered region has Lebesgue measure $m_0+\cdots+m_{k-1}$. Furthermore, during the dispersion of a given mass, the covered domain increases continuously, except when it merges with another covered connected component (CC). We show a very general exchangeability property for the sequence of covered CC. Additionally, we demonstrate a universal spacing property between these CC, and a notable general result: if the $(u_i)$ are independent and rotationally invariant, then the number of free (not covered) CC follows a binomial distribution whose parameters depend solely on the number and total mass of arrived particles. Furthermore, conditional on the number of CC, the sizes of the free CC follow a simple Dirichlet distribution in the continuous case, regardless of the dispersion policy considered and the values of the masses. We also characterize the distribution of the occupied space. In the second part of the paper, we study the total cost associated with these models for various cost models, and establish connections with the additive coalescent. We also provide an asymptotic representation of the limiting covered space as the number of masses goes to infinity.

We introduce a particle model, that we call the $\textit{golf model}$. Initially, on a graph $G$, balls and holes are placed at random on some distinct vertices. The balls then move one by one, doing a random walk on $G$, starting from their initial vertex and stopping at the first empty hole they encounter, which they fill. On finite graphs, under reasonable assumptions (if there are more holes than balls, and if the Markov chain characterizing the random walks is irreducible) a final configuration is reached almost surely. In the paper, we are mainly interested in ${\bf H}^1$, the set of remaining holes. We give the distribution of ${\bf H}^1$ on $\mathbb{Z}/n\mathbb{Z}$, and describe a phase transition for the largest distance between two consecutive holes when the number of remaining holes is of order $\sqrt{n}$. We show that the model on $\mathbb{Z}$ is well-defined if every vertex contains either a ball with probability $d_{\sf b}$, a hole with probability $d_{\sf h}$, or nothing, independently from the other vertices, as long as $d_{\sf b} \leq d_{\sf h}$, and we describe the law of ${\bf H}^1$ in this case.