Negar Mehr

Stable hybrid Model Predictive Control for ramp metering

We formulate the Asymmetric Cell Transmission Model (ACTM) as a piecewise affine system defined over regions of the state and input space. We synthesize a hybrid Model Predictive Controller (MPC) for the piecewise affine model such that persistent feasibility and stability are guaranteed. We do so by designing a terminal constraint and terminal cost for the equilibrium state of the system. We include a detailed analysis of the equilibrium point of the piecewise affine system, and define the region of demand under which an equilibrium point exists. We show that our method achieves the same performance in terms of efficiency and exhibits much smoother behavior than that of the commonly used relaxed ACTM controller formulation.

Inferring and assisting with constraints in shared autonomy

Our goal is to enable robots to better assist people with motor impairments in day-to-day tasks. Currently, such robots are teleoperated, which is tedious. It requires carefully maneuvering the robot by providing input through some interface. This is further complicated because most tasks are filled with constraints, e.g. on how much the end effector can tilt before the glass that the robot is carrying spills. Satisfying these constraints can be difficult or even impossible with the latency, bandwidth, and resolution of the input interface. We seek to make operating these robots more efficient and reduce cognitive load on the operator. Given that manipulation research is not advanced enough to make these robots autonomous in the near term, achieving this goal requires finding aspects of these tasks that are difficult for human operators to achieve, but easy to automate with current capabilities. We propose constraints are the key: maintaining task constraints is the most difficult part of the task for operators, yet it is easy to do autonomously. We introduce a method for inferring constraints from operator input, along with a confidence-based way of assisting the user in maintaining them, and evaluate in a user study.

Optimal mode-switching and control synthesis for floating offshore wind turbines

This paper proposes a multi-objective optimal control and switching strategy for floating offshore wind turbines when the wind speed can be approximately predicted. The system is modeled as a hybrid automaton with two modes corresponding to the turbine operation in low- and high-speed wind profiles. The main control objective in the low-speed wind mode is to maximize the total captured power in a finite time horizon, whereas in the high-speed mode, it is desired to regulate the generator torque and speed around predefined rated values even under gust loads. The problem is formulated as a constrained mixed-integer bilinear program in a model predictive control framework. The posed constraints correspond to the electrical/mechanical limitations of the blade actuators and generators. Various practical considerations, such as minimizing the number of switching occurrences and mechanical fatigue prevention, are explicitly considered in the optimization problem. The proposed control method is applied to the dynamical model of a real wind turbine and simulation results are presented.

Probabilistic controller synthesis for freeway traffic networks

Summary form only given. Traffic management and control in urban environments has been a problem of interest in the recent years due to the costs incurred by congestion delays, green house gas emission, fuel consumption, etc. This would require researchers to design better algorithms capable of maximizing network throughput and performance with the current existing infrastructure. The focus of this work is on designing control strategies that enhance traffic conditions in freeways. An effective control strategy for traffic regulation in freeways is shown to be ramp metering [3]. In order to encode required complex properties for efficient traffic management, control synthesis through temporal logic specifications are proved to be powerful and successful in traffic networks [1,2]. Nonetheless, in all these works, the assumption is that exogenous vehicular demands are known deterministically a priori. This is in contrast to the intrinsic stochastic nature of vehicular demands. In this work, we propose using Signal Temporal Logic (STL) for specifying desired properties in a probabilistic framework allowing for the demands to be treated as random variables. As controlling large scale traffic networks requires macroscopic models with continuous quantities, STL is a perfect candidate as a specification language. Furthermore, STL allows expressing rich temporal properties that encode safety, liveness, response, etc. We assume that the underlying distribution of demands is known and use sampling techniques to optimize for an empirical average of the expected total travel time of the network subject to STL constraints in an MPC fashion.

Stochastic predictive freeway ramp metering from Signal Temporal Logic specifications

We propose a ramp metering strategy capable of treating exogenous arrivals as random variables since freeway network arrivals are stochastic by nature. In order to express desired temporal properties of the network, we adopt Signal Temporal Logic (STL) as our specification language and present a general framework for synthesizing controllers for piecewise affine systems subject to stochastic uncertainties. We synthesize controllers that satisfy these stochastic STL specifications through sample average approximation techniques. We further showcase our approach for a freeway ramp metering example, we use sampling techniques to obtain ramp flows that minimize the expectation of the total travel time.