Research
As James Newman once said, “the theory of groups is a branch of mathematics in which one does something to something and then compares the results with the result of doing the same thing to something else, or something else to the same thing”.
My research is something like this!
Figure: A word-map based on the titles of my publications. The font size of each word shows the frequency of its appearance in the titles (Created by http://www.wordle.net/).
Crowd-Sourcing Estimation with Strategic Senders:
Emerging control applications, such as intelligent transportation systems, rely on networked estimation systems. A typical, yet often implicit, assumption when designing networked estimation systems is that the measurement units are not strategic, i.e., that they directly transmit whatever they measure. In practice, however, this assumption might be unrealistic. For instance, a malicious sender might intercept the stream of data that is being transmitted from the measurement unit and replace it with misleading information. A more subtle reason for dismissing the above mentioned assumption is the possible presence of strategic measurement units in the networked system. To better visualize this scenario, let us employ an example stemming from traffic estimation via crowd-sourcing. In this example, we may distribute an application (designed for smart devices) to many users in a city and ask them to voluntarily report the traffic condition in their vicinity (e.g., Waze). The fused information can then be used to reroute vehicles (through their navigational devices) so as to reduce congestion on some roads. As time passes by, we would inevitably realize that some participants under-report the traffic condition in their neighborhoods (e.g., small business owners might report low traffic in the hope that rerouting brings customers) while some other participants over-report traffic (e.g., residential users might want to repel traffic to reduce noise pollution or to improve safety). Here, we investigate conditions that guarantee the receiver can construct the exact state of the system upon asking enough sender.
For more information, please check my publication page
Collaborators: Prof. Cedric Langbort and Mr. Andre Teixeira.
Decentralized Control Design with Limited Plant Model Information:
In recent years, this has been the main topic of research for me. Let me briefly explain why we are interested in this research problem. Large-scale control systems are often composed of several smaller interconnected units. For these systems, it is common to employ local controllers, which observe and act locally. At the heart of common control design procedures for distributed systems lies the often implicit assumption that the designer has access to the global plant model information when designing a local controller. However, there are several reasons why such plant model information would not be globally known. One reason could be that the designer wants the parameters of each local controller to only depend on local model information, so that the controllers are not modified if the model parameters of a particular subsystem change. It might also be the case that the design of each local controller is done by individual designers with no access to the global plant model, for instance, due to the fact that the designers refuse to share their model information since they consider it private. In this research problem, we investigate the trade-off between the amount of model information exploited by a control design strategy and the best possible closed-loop performance.
For more information, please check my publication page or PhD thesis.
Collaborators: Prof. Karl H. Johansson, Prof. Cedric Langbort, and Prof. Henrik Sandberg.
Distributed Implementation of VCG Mechanisms:
Recently, we started looking at large scale resource allocation problems for multiple agents, and a situation in which agents are suggested to take part in distributed optimization algorithms such as, dual decomposition, to solve the problem. Here, we assumed that the agents are strategic, that is, they want to minimize their own individual cost rather than the global social cost. Therefore, if we do not carefully design a taxing mechanism, these agents are not going to solve the original problem that we were considering. Now, inspired by the classical Vickrey-Clarke-Groves mechanism and more recent algorithmic mechanism design theory, we proposed a tax mechanism that incentivises agents to faithfully implement the intended algorithm. To do so, we defined a new notion of asymptotic incentive compatibility to characterize a desirable property of the proposed class of mechanisms, which provides a sequence of mechanisms that gives agents a diminishing incentive to deviate from suggested algorithm.For more information, please check our recent conference paper here.
Collaborators: Prof. Cedric Langbort and Dr. Takashi Tanaka.
Transportation Networks:
We consider heterogeneous routing and congestion games, where each driver or vehicle belong to a certain type. The type determines the cost of traveling along an edge as function of the flow of all types of drivers or vehicles over that edge. Our interest in solving this problem is motivated by traffic networks in which the assumption that the drivers are of the same type might not be very realistic. For instance, due to fuel consumption, heavy-duty vehicles and cars might experience different costs for using the road even if their travel times are equal. We started examining this phenomenon in congestion games where the heavy-duty vehicles observe an increased efficiency when a higher number of heavy-duty vehicles are present on the same road as them (because of a higher platooning possibility and, therefore, an improved fuel efficiency) while this may not true for cars.
For more information, please check our recent conference paper here.
Collaborators: Prof. Karl H. Johansson, Prof. Alexandre M. Bayen, and Mr. Walid Krichene.
Stochastic Sensor Scheduling:
In this problem, we propose a stochastic sensor scheduling policy with application to networked control systems. We model sampling instances (in a networked system) using jumps between states of a continuous-time Markov chain. We introduce a cost function for this Markov chain which is composed of terms depending on the average sampling frequencies of the subsystems and the effort needed for changing the parameters of the underlying Markov chain. We extract an optimal scheduling policy to fairly allocate network resources (i.e., access to the network) among the control loops. We apply this scheduling policy to a networked control system composed of several scalar decoupled subsystems and compute upper bounds for their closed-loop performance.
For more information, please check our recent conference paper here.
Collaborators: Prof. Karl H. Johansson.