2020高端外专系列学术报告
Topic: Reinforcement Learning for Networked Estimation and Control
Speaker:Prof. Daniel Quevedo(Queensland University of Technology)
Abstract:
The opportunities provided by feedback control of networked dynamical systems are enormous. Yet it is by no means clear
how to harness modern communication, network and computation technologies to obtain high-quality designs. The main
stumbling blocks stem from the significant gaps which exist between understanding of constituent parts and the challenges
faced when bringing them together. Whilst the vast realm of applications of networked cyber-physical systems brings a
variety of issues, a common thread is that many of the standard paradigms that allow the separation of communications
and control systems are no longer valid. Thus, the need for more holistic approaches arises. In this talk, we study the
scheduling of sensor transmissions for wireless estimation and control. We focus on setups where, at each time instant,
a single agent decides which sensors have network access and which ones do not. To address such scheduling problems,
we formulate Markov decision processes and solve them using reinforcement learning techniques.
Arrangement:
Time: 14:00- 15:00, Tuesday, 22nd, September
How to join:
Offline: 公司宝山校区东区机自大楼702(B)会议室
Online: Tencent Meeting & VooV Meeting [腾讯会议]
Link: https://meeting.tencent.com/s/A7xU2n6EaqWi
