On aims and methods of collective animal behaviour

Animal Behaviour, Journal Year: 2024, Volume and Issue: 210, P. 189 - 197

Published: Feb. 27, 2024

Collective animal behaviour is a subfield of behavioural ecology, making extensive use its tools observation, experimental manipulation and model building. However, fundamental ecology approach, the application optimality theory, has been comparatively neglected in collective behaviour. This article seeks to address this imbalance, by outlining an evolutionary theory framework for discipline. The requires number questions be addressed. First, what correct quantity optimize? can achieved via combination considering organisms' life history, alongside such as statistical decision stochastic dynamic programming. Second, mechanism appropriate optimal behaviour? involves ensuring that models are self-consistent rather than assuming parameter values. Third, at level selection does optimization act? Selection acts on individual except very particular circumstances, yet phenomena group level, thus introducing risk confusing adaptive properties emerge. presents examples under each three questions, well discussing mismatches between observation. In doing so, it hoped fully inherits philosophy parent discipline ecology.

Language: Английский

---

Dynamics of an information theoretic analog of two masses on a spring

Chaos Solitons & Fractals, Journal Year: 2024, Volume and Issue: 188, P. 115535 - 115535

Published: Sept. 17, 2024

Language: Английский

---

Body orientation change of neighbors leads to scale-free correlation in collective motion

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Oct. 17, 2024

Collective motion, such as milling, flocking, and collective turning, is a common captivating phenomenon in nature, which arises group of many self-propelled individuals using local interaction mechanisms. Recently, vision-based mechanisms, establish the relationship between visual inputs motion decisions, have been applied to model better understand emergence motion. However, previous studies often characterize input transient Boolean-like sensory stream, makes it challenging capture salient movements neighbors. This further hinders onset response mechanisms increases demands on sensing devices robotic swarms. An explicit context-related cue serving for decision-making still lacking. Here, we hypothesize that body orientation change (BOC) significant characterizing salience neighbors, facilitating response. To test our hypothesis, reveal role BOC during U-turn behaviors fish schools by reconstructing scenes from view individual fish. We find an with larger takes leading U-turns. explore this empirical finding, build pairwise mechanism basis BOC. Then, conduct experiments spin turn real-time physics simulator investigate dynamics information transfer BOC-based validate its effectiveness 50 real miniature swarm robots. The experimental results show not only facilitates directional within but also leads scale-free correlation swarm. Our study highlights practicability governed neighbor's robotics effect enhancing

Language: Английский

---

CBIL: Collective Behavior Imitation Learning for Fish from Real Videos

ACM Transactions on Graphics, Journal Year: 2024, Volume and Issue: 43(6), P. 1 - 17

Published: Nov. 19, 2024

Reproducing realistic collective behaviors presents a captivating yet formidable challenge. Traditional rule-based methods rely on hand-crafted principles, limiting motion diversity and realism in generated behaviors. Recent imitation learning learn from data but often require ground-truth trajectories struggle with authenticity, especially high-density groups erratic movements. In this paper, we present scalable approach, Collective Behavior Imitation Learning (CBIL), for fish schooling behavior directly videos , without relying captured trajectories. Our method first leverages Video Representation Learning, which Masked AutoEncoder (MVAE) extracts implicit states video inputs self-supervised manner. The MVAE effectively maps 2D observations to that are compact expressive following the stage. Then, propose novel adversarial capture complex movements of schools fish, enabling efficient distribution patterns measured latent space. It also incorporates bio-inspired rewards alongside priors regularize stabilize training. Once trained, CBIL can be used various animation tasks learned priors. We further show its effectiveness across different species. Finally, demonstrate application our system detecting abnormal in-the-wild videos.

Language: Английский

---

Computational joint action: Dynamical models to understand the development of joint coordination

PLoS Computational Biology, Journal Year: 2024, Volume and Issue: 20(10), P. e1011948 - e1011948

Published: Oct. 22, 2024

Coordinating with others is part of our everyday experience. Previous studies using sensorimotor coordination games suggest that human dyads develop strategies can be interpreted as Nash equilibria. However, if the players are uncertain about what their partner doing, they which robust to actual partner's actions. This has suggested humans select actions based on an explicit prediction will doing-a model-which probabilistic by nature. mechanisms underlying development a joint over repeated trials remain unknown. Very much like adaptation individuals external perturbations (eg force fields or visual rotations), dynamical models may help understand how develops trials. Here we present general computational model-based game theory and Bayesian estimation-designed Joint tasks modeled quadratic games, where each participant's task expressed cost function. Each participant predicts next move (partner model) optimally combining predictions sensory observations, selects through stochastic optimization its expected cost, given model. The model parameters include perceptual uncertainty (sensory noise), representation (retention rate internale in action selection decay (which action's learning rate). used two ways: (i) simulate interactive behaviors, thus helping make specific context scenario; (ii) analyze time series experiments, providing quantitative metrics describe individual behaviors during action. We demonstrate variety scenarios. In version Stag Hunt game, different representations lead via-point (2-VP) reaching task, consist complex trajectories, captures well observed temporal evolution performance. For this also estimated from experimental provided comprehensive characterization dyad participants. Computational identifying factors preventing facilitating coordination. They clinical settings, interpret impaired interaction capabilities. provide theoretical basis devise artificial agents establish forms facilitate neuromotor recovery.

Language: Английский

---