A simple multi-agent system can be effectively utilized in disaster response applications, such as firefighting. Such a swarm is required to operate in complex environments with limited local sensing and no reliable inter-agent communication or centralized control. These simple robotic agents, also known as Ant Robots, are defined as anonymous agents that possess limited sensing capabilities, lack a shared coordinate system, and do not communicate explicitly with one another. A key challenge for simple swarms lies in maintaining cohesion and avoiding fragmentation despite limited-range sensing. Recent advances in machine learning offer effective solutions to some of the classical decentralized control challenges. We propose VariAntNet, a deep learning-based decentralized control model designed to facilitate agent swarming and collaborative task execution. VariAntNet includes geometric features extraction from unordered, variable-sized local observations. It incorporates a neural network architecture trained with a novel, differentiable, multi-objective, mathematically justified loss function that promotes swarm cohesiveness by utilizing the properties of the visibility graph Laplacian matrix. VariAntNet is demonstrated on the fundamental multi-agent gathering task, where agents with bearing-only and limited-range sensing must gather at some location. VariAntNet significantly outperforms an existing analytical solution, achieving more than double the convergence rate while maintaining high swarm connectivity across varying swarm sizes. While the analytical solution guarantees cohesion, it is often too slow in practice. In time-critical scenarios, such as emergency response operations where lives are at risk, slower analytical methods are impractical and justify the loss of some agents within the swarm. This paper presents and analyzes this trade-off in detail.

A metamorphic robotic system (MRS) consists of anonymous modules, each of which autonomously moves in the 2D square grid by sliding and rotation with keeping connectivity among the modules. Existing literature considers distributed coordination among modules so that they collectively form a single MRS. In this paper, we consider distributed coordination for two MRSs. We first present a rendezvous algorithm that makes the two MRSs gather so that each module can observe all the other modules. Then, we present a merge algorithm that makes the two MRSs assemble and establish connectivity after rendezvous is finished. These two algorithms assume that each MRS consists of five modules, that do not have a common coordinate system. Finally, we show that five modules for each MRS is necessary to solve the rendezvous problem. To the best of our knowledge, our result is the first result on distributed coordination of multiple MRSs.

Patrolling is a key problem in robotics wherein a mobile agent or team of mobile agents are tasked with repeatedly visiting every vertex of a graph environment by traversing edges. This task has well-known applications to navigation, web crawling, and swarm intelligence [19]. Patrolling has been studied under diverse sets of assumptions regarding e.g. the number of agents, the capabilities of each agent, and the underlying graph environment [22]. A central question related to patrolling is the space complexity of patrolling an environment: the amount of memory required by the agent(s) to patrol the environment [5, 8, 11, 13, 15, 18, 22]. The simplest kind of environment that is interesting to patrol is perhaps the grid graph. Hence, the goal of this work is to establish the space complexity of patrolling d-dimensional grid graphs with a single mobile agent that has limited visibility. More concretely, suppose a mobile agent with fixed orientation, R, is placed somewhere inside a grid graph. We assume R has b bits of internal state memory and the ability to see locations at Manhattan distance V or less from itself (see Figure 1), and must act based only on this information, i.e., it is a finite automaton. For what values of b and V does there exist an algorithm that enables R to patrol the grid?

Multipath Generalized Adaptive A* (MPGAA*) is an A*- based incremental search algorithm for dynamic terrain that can outperform D* for the (realistic) case of limited visibility ranges. A first contribution of this paper is a brief analysis studying why MPGAA* has poor performance for extended visibility ranges, which concludes that MPGAA* carries out an excessive number of heuristic updates. Our second contribution is a method to reduce the number of heuristic updates that preserves optimality. Finally, a third contribution is a variant of MPGAA*, MPGAA*-back, which we show outperforms MPGAA* and D* on a wide range of dynamic grid pathfinding scenarios, and visibility ranges.