权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

NRI: INT: COLLAB: Synergetic Drone Delivery Network in Metropolis

NRI：INT：COLLAB：大都市的协同无人机交付网络

基本信息

批准号：
1830639
负责人：
Naira Hovakimyan
金额：
$ 103.77万
依托单位：
University of Illinois at Urbana-Champaign
依托单位国家：
美国
项目类别：
Standard Grant
财政年份：
2018
资助国家：
美国
起止时间：
2018-09-15 至 2022-08-31
项目状态：
已结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=1830639&HistoricalAwards=false
关键词：
NRI INT COLLAB Synergetic Drone

项目摘要

Synergetic Drone Delivery Network in MetropolisThe rapid growth of e-commerce demands has put additional strain on dense urban communities resulting in increased traffic of delivery trucks while slowing down the pace of delivery operations. With recent quick-purchase innovations like the Amazon Dash button, e-commerce drastically modified the consumers behavior to buy smaller products separately and regularly, adding more load to delivery operations. Another growing trend is the offering of fast delivery services such as same-day and instant delivery. Instacart, Uber Eats and Amazon Now are examples of services that can fulfill a delivery order in just under 2 hours. These services rely heavily on the infrastructure of ride-sharing vehicles as Uber or Lyft drivers. This solution offers great flexibility to the consumer, but a single person can only deliver one purchase order to a customer at a time, and it is not scalable or cost-effective. There is an unquestionable need to redesign the current method of distribution packages in urban environments. This project envisions a framework that synergizes manipulatable distribution networks, comprising autonomous flying robots (drones) with existing transport networks, towards enhanced autonomy and economics in logistics. Imagine that a ride-sharing vehicle outfitted with a docking device for packages on its roof is traveling through a distribution center towards downtown. A drone can place a package on the vehicle's roof while it drives by the distribution center, and another drone can recover the package once the vehicle is driving through another distribution center in proximity to its destination. An operator that owns several base stations, at each of which it employs a network of drones to pick packages from the respective base station and drop it on a ground vehicle assigned to the package, is a required assumption by the framework. The ground vehicles can be public transport vehicles (PTVs), ride-sharing vehicles (RSVs), or operator owned vehicles (OOVs), which carry the package for most of the distance.The approach relies on three main thrusts: i) socially aware robotics, ii) safe and robust motion planning and execution, iii) cooperative network logistics. Motion planning for robots will be developed with account of peoples perception of safety, privacy, and comfort. Socially-aware motion planning methods to generate trajectories with guarantees of safety in the presence of obstacles and humans will be developed. Psychological experiments will be developed to study human's subtle behavior in response to the presence of multiple drones using virtual reality test environment. Local control algorithms will be developed for each drone to follow a feasible collision free path. Robust local communication protocols will be investigated so that flying robots can perform collaborative tasks over busy air/ground traffic conditions and unreliable communication networks. Another objective is to achieve robust and safe rendezvous with fast moving vehicles under communication, schedule, and other modeling uncertainties. Algorithms that generate (possibly multi-hop) routes for each package, consisting of vehicle route segments, with the objective of minimizing cumulative delivery time, will be developed. The series of vehicle segments on which each package travels, and the associated schedule, is required as input for drones. This in turn necessitates solving the underlying network design problem for the centralized entity, to determine locations of distribution centers (bases) and number of OOVs required for feasible and reliable delivery of all packages, while explicitly estimating uncertainty from traffic trends and overall frequency of travel of RSVs between various points in the network. Game-theoretic mechanisms that incentivize cooperation among multiple independent operators of PSVs and RSVs will be developed. Mechanisms have to be specifically designed to ensure truthful bidding, because the objectives of the operator, the RSVs and the PSVs are not naturally aligned.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

大都市协同无人机配送网络电子商务需求的快速增长给人口密集的城市社区带来了额外的压力，导致配送卡车的交通量增加，同时减缓了配送业务的速度。随着亚马逊快捷键（Amazon Dash button）等最近的快速购买创新，电子商务极大地改变了消费者的行为，使他们能够单独、定期地购买较小的产品，这给配送业务增加了更多的负担。另一个日益增长的趋势是提供快速送货服务，如当日送达和即时送达。Instacart、Uber Eats和Amazon Now都是可以在两小时内完成配送订单的服务。这些服务严重依赖乘车共享车辆的基础设施，如优步或Lyft司机。这种解决方案为消费者提供了很大的灵活性，但是一个人一次只能向一个客户交付一个采购订单，并且它不具有可伸缩性或成本效益。毫无疑问，需要重新设计目前在城市环境中分发包的方法。该项目设想了一个框架，将可操纵的分销网络协同起来，包括自主飞行机器人（无人机）和现有的运输网络，以增强物流的自主性和经济性。想象一下，一辆在车顶安装了包裹对接装置的拼车正穿过一个配送中心驶向市中心。当车辆经过配送中心时，无人机可以将包裹放在车顶上，当车辆经过目的地附近的另一个配送中心时，另一架无人机可以收回包裹。一个运营商拥有多个基站，每个基站都有一个无人机网络，从各自的基站取包裹，并将其投放到分配给包裹的地面车辆上，这是该框架所要求的假设。地面车辆可以是公共交通车辆（ptv）、拼车车辆（rsv）或运营商拥有的车辆（oov），它们可以在大部分距离内携带包裹。该方法依赖于三个主要推动力：i)社会感知机器人，ii)安全和稳健的运动规划和执行，iii)合作网络物流。机器人的运动规划将随着人们对安全、隐私和舒适的感知而发展。将开发具有社会意识的运动规划方法，以生成在障碍物和人类存在的情况下保证安全的轨迹。将开发心理实验，利用虚拟现实测试环境研究人类对多架无人机存在的微妙行为。将为每架无人机开发局部控制算法，使其遵循可行的无碰撞路径。鲁棒的本地通信协议将被研究，以便飞行机器人可以在繁忙的空中/地面交通条件和不可靠的通信网络中执行协作任务。另一个目标是在通信、调度和其他建模不确定的情况下实现与快速移动车辆的鲁棒安全交会。将开发为每个包裹生成（可能是多跳）路线的算法，这些路线由车辆路线段组成，目标是最小化累积交付时间。每个包裹所在的一系列车辆分段，以及相关的时间表，都需要作为无人机的输入。这反过来又需要解决集中式实体的潜在网络设计问题，以确定配送中心（基地）的位置和可行且可靠地交付所有包裹所需的oov数量，同时明确估计流量趋势和rsv在网络中各点之间的总体旅行频率的不确定性。将开发激励psv和rsv的多个独立运营商之间合作的博弈论机制。由于运营商、rsv和psv的目标并非自然一致，因此必须专门设计机制以确保真实竞标。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

期刊论文数量（19）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

Learning to Communicate: A Machine Learning Framework for Heterogeneous Multi-Agent Robotic Systems

DOI：
10.2514/6.2019-1456
发表时间：
2018-12
期刊：
ArXiv
影响因子：
0
作者：
Hyung-Jin Yoon;Huaiyu Chen;Kehan Long;Heling Zhang;Aditya Gahlawat;Donghwan Lee;N. Hovakimyan
通讯作者：
Hyung-Jin Yoon;Huaiyu Chen;Kehan Long;Heling Zhang;Aditya Gahlawat;Donghwan Lee;N. Hovakimyan

Improving the Robustness of Reinforcement Learning Policies With ${\mathcal {L}_{1}}$ Adaptive Control