Collective control as a decentralized task allocation testbed

semanticscholar(2021)

引用 0|浏览1
暂无评分
摘要
We present a collective tracking problem in which the agents in a swarm collectively control a tracker to follow a target moving in two dimensional space. Tasks are represented by the four cardinal directions and task demand refers to differences between the target and tracker locations in each of the directions. In any given timestep, each agent can choose to address task demand in a single direction. Tracker movement in a timestep is generated by aggregating the choices of all of the agents in the swarm in that timestep. Thus, the agents in a swarm collectively move a tracker in response to a target path. This problem is representative of a general decentralized task allocation problem in which one or more tasks exhibit demand that is addressed by allocating an appropriate number of agents to each task. The tracking problem allows us to define dynamically changing task demands in a systematic way in the form of target paths. Because the variety of target paths that can be defined is essentially infinite, this problem allows us to define task allocation problems over a wide range of characteristics requiring different responses from the swarm. In addition to quantitative evaluation metrics, the actual two dimensional target and tracker paths provide an intuitive way to visualize system performance.
更多
查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要