ABSTRACT: The first part of this talk is motivated by the fact that learning algorithms are growing in popularity for sequential decision making in many cyber-physical systems. However, when dealing with safety-critical systems, it is paramount that the learner’s actions do not violate the safety/reliability constraints of the system at any round, in spite of uncertainty about system parameters. An example we will highlight is that of optimal real-time price design for demand management in power distribution systems given unknown customer price response functions. We will showcase the performance of a “safety-aware” bandit heuristic for designing prices that controls the probability of violation of power grid constraints during the learning process. We then study the effect of such safety constraints on the growth of regret for certain classes of stochastic bandit optimization problems.

In the second part of the talk, we consider the problem of joint routing, battery charging, and pricing problem faced by a profit-maximizing transportation service provider that operates a fleet of autonomous electric vehicles. To accommodate for the time-varying nature of trip demands, renewable energy availability, and electricity prices and to further optimally manage the autonomous fleet, a dynamic pricing and control policy is required. We highlight several such policies, including one trained through deep reinforcement learning to develop a near-optimal control policy. We also determine the optimal static policy to serve as a baseline for comparison with our dynamic policy and for determining the capacity region of the system. While the static policy provides important insights on optimal pricing and fleet management, we show that in a real dynamic setting, it is inefficient to utilize a static policy.

BIO: Mahnoosh Alizadeh is an assistant professor of Electrical and Computer Engineering at the University of California Santa Barbara. She received the B.Sc. degree in Electrical Engineering from Sharif University of Technology in 2009 and the M.Sc. and Ph.D. degrees from the University of California Davis in 2013 and 2014 respectively, both in Electrical and Computer Engineering. From 2014 to 2016, she was a postdoctoral scholar at Stanford University. Her research is focused on the design of network control and optimization algorithms for societal-scale cyber-physical systems, with a particular focus on renewable energy integration in the power grid and electric transportation systems. She is a recipient of the NSF CAREER award.

Hosted by: Ashutosh Nayyar