SEI: Collaborative Research: Computational Tools for Managing Herds

本项目的目标是开发用于研究具有自然移动性和社会交互作用的智能体群体的计算方法。由于生物和非生物因素之间的复杂相互作用，这类系统中的代理人可以自行移动，最终形成可以使用物理数据进行研究和建模并最终受到控制的行为。该小组将开发算法，用于现场试验期间的自适应数据收集，试验后数据分析，以建模行为，并使用现场试验中学习的模型和来自卫星图像的环境数据来控制群体的运动。这项工作的主要动机和应用是在自由放养动物生态领域。畜群等动物群体是复杂的自组织系统，受年龄、群体大小、景观地形、植物物候、天气和环境等多种因素的影响。该项目结合了计算机网络、建模和机器人运动规划和控制方面的最新进展。该团队将在野外收集大量时间的物理数据，以定义动物群体的行为模型，使用这些模型来开发规划和控制算法，以协调这些群体的位置，并使用位置控制系统来管理种群密度和优化土地利用。这包括：(1)跟踪单个动物的运动和活动；(2)识别自然群体的形成和行为；(3)计划和控制个体和群体的位置；(4)管理群体的密度；以及(5)使用动物群体作为联网的信息骨干。尽管这些模型将基于特定的领域数据，但建议的方法和由此产生的算法将适用于其他自由放养动物协会和人工合成动物协会。这项工作将把计算机科学、机器人学和动物专家聚集在一个基于这些领域之间协同作用的项目中。这项工作将基于真实数据，由此产生的模型和算法将用于定义野外实验的详细计划。这些模型将导致更好地理解动物行为和个体和群体层面的控制，这可能会对农业和环境产生广泛影响。这项研究将开发控制放牧密度的方法，并将觅食变成通过控制放牧模式来修复牧场生态系统的实用工具。这项研究将有助于培养新一代学生与物理世界相互作用的计算能力。

ABSTRACTThe goal of this project is to develop computational approaches for studying groups of agents with natural mobility and social interactions. The agents in such systems can move on their own due to complex interactions between biotic and abiotic factors, culminating in behaviors that can be studied and modeled using physical data, and ultimately controlled. The team will develop algorithms for adaptive data collection during a field experiment,post-experiment data analysis for modeling behavior, and control of the movement of the group using the learned models in field experimentsand environmnet data from satellite imagery. The main motivation and application of this work is in the area of free-ranging animal ecology. Groups of animals such as herds are complex self-organized systems that are affected by many factors including age, group size, landscape topography, plant phenology, weather, and the environment.This project combines the most recent advances in computer networking, modeling, and robot motion planning and control. The team will collect physical data for extensive periods of time in the field to define behavior models for herds of animals, using these models to develop planning and control algorithms for coordinating the location of these herds, and using the location control system to manage the stock density and optimize the use of land. This includes: (1) tracking the motion and activities of individual animals; (2) identifying natural group formation and behavior; (3) planning and controlling the locationof individuals and groups; (4) managing the density of the group; and (5) using animal groups as networked information backbones. Although the models will be grounded in specific field data, the proposed methodology and resulting algorithms will be applicable to other societies of free-ranging animals and synthetic societies of agents.This work will bring together computer science, robotics, and animalexperts in a project that rests on the synergy between these domains.The work will be grounded on real data and the resulting modelsand algorithms will be used to define detailed plans for field experiments.The models will lead to a better understanding of animal behavior and control at the individual and group level, which has the potential to impact broadly agriculture and the environment. This research will develop methods to manipulate stocking density and turn foraging into a practical tool to remediate range ecosystems by controlling grazing patterns.This research will contribute to training of a new generation ofstudents in computation for interaction with the physical world.