Abstract: In human-robot interaction or more generally multi-agent systems, we often have decentralized agents that need to perform a task together. In such settings, it is crucial to have the ability to anticipate the actions of other agents. Without this ability, the agents are often doomed to perform very poorly. Humans are usually good at this, and it is mostly because we can have good estimates of what other agents are trying to do. We want to give such an ability to robots through reward learning and partner modeling. In this talk, I am going to talk about active learning approaches to this problem and how we can leverage preference data to learn objectives. I am going to show how preferences can help reward learning in the settings where demonstration data may fail, and how partner-modeling enables decentralized agents to cooperate efficiently.

Bio: Dr. Erdem Bıyık is a postdoctoral researcher at the Center for Human-Compatible Artificial Intelligence at University of California, Berkeley. He has received his B.Sc. degree from Bilkent University, Turkey, in 2017; and Ph.D. degree from Stanford University in 2022. His research interests lie in the intersection of robotics, artificial intelligence, machine learning and game theory. He is interested in enabling robots to actively learn from various forms of human feedback and designing robot policies to improve the efficiency of multi-agent systems both in cooperative and competitive settings. He also worked at Google as a research intern in 2021 where he adapted his active robot learning algorithms to recommender systems. He will join the University of Southern California as an assistant professor in 2023.

Abstract: The low-cost Internet of Things architecture is well suited for remote environmental monitoring. This talk covers a device designed to be low cost, solar powered, and used outdoors, for critical stream level monitoring in the drought stricken North Bay. The “open source” Stroud Water Research Center's "Modular Sensors" software and the Mayfly microcomputer facilitate environmental scientists collecting physical measurements from the “great outdoors”. Low-cost board level sensors using ADC, I2C or 1Wire are well understood and relatively low power. External transducers have a wider range of sensing, and require interfaces using the newer USGS SDI-12 three wire water world protocol, and the older stable Modbus 4wire RS485. Each brings new powering challenges. The software code and hardware definitions are stored in git, a distributed “Version Control System” allowing easy incremental traceable improvements. The package comes with technical debt that needs characterizing and analyzing.

Bio: Neil Hancock is a Chartered Engineer, active in circuit design and firmware for 40 years. For the first 20 years he worked with large companies in telecom equipment design and more recently in Internet of Things design for environmental monitoring. Neil volunteers as a Planning Commissioner for the City of Cotati, an Industry Advisor for SSU, is a long-term IEEE member, and enjoys short bicycle rides to coffee shops in Sonoma County. He received his BSc in Electronic Engineering, from City, University of London, United Kingdom.

Chartered Engineers (CEng) develop solutions to engineering problems using new or existing technologies, through innovation, creativity and change and/or they may have technical accountability for complex systems with significant levels of risk.