Robots Organizing Environments: Collective Strategies based on Low-Cost Navigation

人们常说机器人最适合干肮脏、枯燥和危险的工作。 该研究计划涉及技术的发展，使一群机器人能够组织他们的环境，这项任务很容易被描述为肮脏，枯燥和潜在的危险。 我们专注于对象聚类和排序的问题。 在对象聚类中，环境中只有一种类型的对象，目标是将所有对象聚集到一个簇中。 在对象排序中，有多种类型的对象，目标是形成同质聚类，理想情况下，每种类型一个。 这项工作属于福尔斯域的群体机器人，这是一种方法来设计的多机器人系统，其中的机器人被约束到本地感测和行动。 在群机器人系统中，没有单个机器人负责，这也意味着任何单个机器人的故障不会导致系统的故障。 以前的对象聚类和排序方法只处理形状一致的对象，这些对象很容易识别。 我们有兴趣解决现实世界的挑战，如不容易识别的任意形状的物体。 另一个现实世界的挑战是处理机器人变得过度拥挤和生产力低下的拥堵。 与人类用户打交道并利用他们的反馈来修改群体的行为是我们将解决的另一个重大现实挑战。我们的方法是创新和实用的，因为我们专注于廉价机器人的能力，可以作为一个集体执行这些任务。 具有所需功能的机器人可以用当前的现成技术建造，价格不到500美元。 我们称之为低成本导航的关键能力包括视觉寻的，使用视觉返回到以前访问过的地方的能力，以及里程计，估计短距离移动的能力。 我们将利用导航策略，通过让机器人故意移动到以前访问过的地方来加速任务性能，而不是目前流行的随机运动策略。将低成本导航能力引入群体机器人将有助于推动该领域走向实际应用。 我们设想机器人清洁房屋，收集花园垃圾，分类可回收物品，并为混乱的工作环境带来秩序。 这些任务可能被视为平凡，但这正是我们对自动化感兴趣的原因。 它们也无处不在。 因此，可行的解决方案有可能产生重大的社会影响，制定这种解决方案有能力产生重大的经济活动和就业机会。 在该计划中接受培训的HQP将有独特的机会利用该研究计划的技术来创建初创公司或在不断发展的机器人行业中发展职业生涯。

It is often stated that robots are best suited for dirty, dull, and dangerous jobs. This research program concerns the development of techniques that would allow a swarm of robots to organize their environment, a task that can easily be described as dirty, dull, and potentially dangerous. We focus on the problems of object clustering and sorting. In object clustering, there is only one type of of object in the environment and the goal is to gather all objects into one cluster. In object sorting, there are multiple types of objects and the goal is to form homogeneous clusters, ideally one for each type. This work falls under the domain of swarm robotics which is an approach to the design of multi-robot systems where the robots are constrained to sense and act locally. In a swarm robotic system, no single robot is in charge which also means that the failure of any single robot does not lead to the failure of the system. Previous approaches to object clustering and sorting only handle uniformly shaped objects which are easily identifiable. We are interested in addressing real-world challenges such as arbitrarily shaped objects which are not readily identifiable. Another real-world challenge is dealing with congestion where the robots become overcrowded and unproductive. Dealing with human users and using their feedback to modify the behaviour of a swarm is another significant real-world challenge that we will address.******Our approach is innovative and practical in that we focus on the capabilities of inexpensive robots which can perform these tasks as a collective. Robots with the desired capabilities can be constructed with current off-the-shelf technology for less than $500. The key capabilities for what we term low-cost navigation include visual homing, the ability to return to previously visited places using vision, and odometry, the ability to estimate movement over short distances. We will exploit strategies for navigation to accelerate task performance by having robots move deliberately to previously visited places, as opposed to the currently popular strategy of randomized motion.******The introduction of low-cost navigation capability into swarm robotics will help to push the field towards practical application. We envision robots cleaning homes, gathering garden waste, sorting recyclables, and bringing order to chaotic work environments. These tasks may be viewed as mundane, but that is exactly why we are interested in automating them. They are also ubiquitous. Therefore, workable solutions have the potential for a major societal impact and the development of such solutions has the capacity to generate significant economic activity and jobs. The HQP trained in this program will have the unique opportunity to exploit the technology from this research program to either create start-up companies or develop careers in the growing robotics industry.