arXiv subjects

Andrea Lama

Publications and source records attributed to Andrea Lama.

5 recordsLinked to original sources

arXiv2025

Hierarchical Learning-Based Control for Multi-Agent Shepherding of Stochastic Autonomous Agents

Multi-agent shepherding represents a challenging distributed control problem where herder agents must coordinate to guide independently moving targets to desired spatial configurations. Most existing control strategies assume cohesive target behavior, which frequently fails in practical applications where targets exhibit stochastic autonomous behavior. This paper presents a hierarchical learning-based control architecture that decomposes the shepherding problem into a high-level decision-making module and a low-level motion control component. The proposed distributed control system synthesizes effective control policies directly from closed-loop experience without requiring explicit inter-agent communication or prior knowledge of target dynamics. The decentralized architecture achieves cooperative control behavior through emergent coordination without centralized supervision. Experimental validation demonstrates superior closed-loop performance compared to state-of-the-art heuristic control methods, achieving 100\% success rates with improved settling times and control efficiency. The control architecture scales beyond its design conditions, adapts to time-varying goal regions, and demonstrates practical implementation feasibility through real-time experiments on the Robotarium platform.

eess.SY↗

arXiv2025

Nonreciprocal field theory for decision-making in multi-agent control systems

Field theories for complex systems traditionally focus on collective behaviors emerging from simple, reciprocal pairwise interaction rules. However, many natural and artificial systems exhibit behaviors driven by microscopic decision-making processes that introduce both nonreciprocity and many-body interactions, challenging these conventional approaches. We develop a theoretical framework to incorporate decision-making into field theories using the shepherding control problem as a paradigmatic example of a multi-agent control system, where agents (herders) must coordinate to confine another group of agents (targets) within a prescribed region. By introducing continuous approximations of two key decision-making elements - target selection and trajectory planning - we derive field equations that capture the essential features of this distributed control problem. Our theory reveals that different decision-making strategies emerge at the continuum level, from average attraction to highly selective choices, and from undirected to goal-oriented motion, driving transitions between homogeneous and confined configurations. The resulting nonreciprocal field theory not only describes the shepherding problem but provides a general framework for incorporating decision-making into continuum theories of collective behavior, with implications for applications ranging from robotic swarms to crowd management systems.

cond-mat.soft↗

arXiv2024

Emergent Cooperative Strategies for Multi-Agent Shepherding via Reinforcement Learning

We present a decentralized reinforcement learning (RL) approach to address the multi-agent shepherding control problem, departing from the conventional assumption of cohesive target groups. Our two-layer control architecture consists of a low-level controller that guides each herder to contain a specific target within a goal region, while a high-level layer dynamically selects from multiple targets the one an herder should aim at corralling and containing. Cooperation emerges naturally, as herders autonomously choose distinct targets to expedite task completion. We further extend this approach to large-scale systems, where each herder applies a shared policy, trained with few agents, while managing a fixed subset of agents.

eess.SY↗

arXiv2024

Shepherding control and herdability in complex multiagent systems

We study the shepherding control problem where a group of "herders" need to orchestrate their collective behaviour in order to steer the dynamics of a group of "target" agents towards a desired goal. We relax the strong assumptions of targets showing cohesive collective behavior in the absence of the herders, and herders owning global sensing capabilities. We find scaling laws linking the number of targets and minimum herders needed, and we unveil the existence of a critical threshold of the density of the targets, below which the number of herders needed for success significantly increases. We explain the existence of such a threshold in terms of the percolation of a suitably defined herdability graph and support our numerical evidence by deriving and analysing a PDE describing the herders dynamics in a simplified one-dimensional setting. Extensive numerical experiments validate our methodology.

cond-mat.stat-mech↗

arXiv2023

Preliminary results on the herdability of complex multiagent systems via local information

We present preliminary results on the problem of driving the dynamics of a group of agents, the herders, so as to steer the collective behaviour of another group of agents, the targets, interacting with them. We define this problem as the multiagent herding control problem and study the herdability of the target agents by relaxing some strong assumptions often made in the existing literature, namely that targets are cohesive in the absence of herders and that herders possess global information. We find that scaling laws exist linking crucial parameters such as the herders' sensing radius and the targets density with the herdability of the system. Our most surprising observation, supported by exhaustive numerical experiments, is that the crucial effect of limited sensing is that the minimum number of herders required to solve the problem significantly increases below a critical minimum density of the target agents.

eess.SY↗