Spatiotemporal Relationship Aided Adaptive Collaboration for Resource Constrained Swarms
Lecture & Organization
Department of ECE, Carnegie Mellon University, USA
Date & Time
Sat 6/6/2020 11:00-12:00am
Zoom ID: 574 459 9066
Dr. Xinlei Chen is a postdoctoral research associate in Department of Electrical and Computer Engineering at Carnegie Mellon University. He received his Bachelor's and Master’s degree in Electronic Engineering from Tsinghua University and his Ph.D. degree in Electrical and Computer Engineering from Carnegie Mellon University. His research interests include large scale cyber physical systems, Artificial Intelligence of Things (AIoT), swarm intelligence, and ubiquitous computing. He has published in both top-tier ACM conferences (eg. SenSys, UbiComp) and high-impact ACM/IEEE journals (eg. IEEE TMC, IEEE JSAC). He has been the recipient of awards including Best Demo Award in ACM SenSys 2016, Best Poster Runner Up Award in ACM SenSys 2016, and Best Poster Award in IEEE/ACM IPSN 2017. His research and demo systems were reported by many well-known media including NBC News, PC World and etc.
Swarm intelligence, as a game-changing technology, has been regarded as an essential component of artificial intelligence, which has broad prospects for both military and civilian applications. For example, it can be applied on unmanned micro aerial vehicles (MAVs) to perform 4D (dull, dirty dangerous and deep) tasks, such as investigation, detection, projection and etc. In addition, swarm intelligence can also be adopted for urban surveillance, radiation monitoring, search & rescue, etc. Due to constraints from real application scenarios, all/most nodes in the swarm have access to limited resources in terms of equipped hardware and acquired data. Therefore, it is challenging for the swarm to achieve satisfying systematic performance with limited resources in real application scenarios, which are hostile, dynamic and complicated. My research focuses on optimizing systematic performances of status estimation, environment inferring and task planning for resource constrained swarms. The spatiotemporal relationships among individual nodes and the acquired data are extracted to improve the entire swarm performance, which is also aided by adaptive actuation strategies based on sensing and learning results. Supported by funding from NSF, DARPA and companies (Intel, Nokia, etc.), my works have been implemented and evaluated on a wide range of intelligent swarm systems: 1) indoor navigation and deployment with MAV swarms; 2) city scale fine grained air pollution inferring with vehicular sensing platforms; 3) city-scale route actuation on ride sharing vehicular mobile crowdsensing platform.