ARC Researchers at the ASME
2017 Dynamic Systems and Control Conference
(October 11-13, 2017 at Tysons Corner, Virginia)

ARC researchers (principal investigators in bold) will be presenting their latest research developments. Below are their papers which may include non-ARC funded research (paper titles in bold are ARC funded).

32-1 Control of Smart Buildings and Microgrids
Wednesday, October, 11 04:00 PM - 06:00 PM
DSCC2017-5362 A Communication-Free Decentralized Power Control Approach for Power Loss Minimization in Islanded Microgrids Using Extremum Seeking
Mr. Su-Yang Shieh, Research Asistent, Univ. of Michigan
Dr. Tulga Ersal, Assistant Research Scientist, Univ. of Michigan
Prof. Huei Peng, Professor, Univ. of Michigan
(not ARC funded)
Abstract: This paper considers islanded microgrids and is motivated by the need for decentralized control strategies with minimal communication among grid components to support a robust and plug-and-play operation. We focus on the problem of power allocation among the distributed generation units (DGs) to maintain low distribution power loss in the grid and develop a communication-free distributed power control approach for power loss minimization based on the extremum-seeking (ES) method. In this approach, the DGs implement ES simultaneously and separately to minimize their current outputs by controlling the active power. The total power loss is thus reduced and no grid structure information or communication is needed in the optimization process. The existence of a Nash equilibrium in the resulting non-cooperative game is proved. Numerical simulations are conducted to demonstrate the performance of the proposed communication-free power control approach and show that it is suitable for maintaining low power loss under different operating conditions in a plug-and-play manner.
Session 5-1 Uncertain Systems and Robustness I
Wednesday, October 11, 04:00 PM - 06:00 PM
DSCC2017-5223 Robustness and Adaptability Analysis of Future Military Modular Fleet Operation System
Mr. Xingyu Li, Research Assistant, Univ. of Michigan
Prof. Bogdan Epureanu, Assistant Professor, Univ. Of Michigan
Abstract: Modular vehicles are vehicles with interchangeable substantial components also known as modules. Fleet modularity provides a system with extra operational flexibility through on-field actions, in terms of vehicle assembly, disassembly, and reconfiguration. The ease of assembly and disassembly of modular vehicles enables them to achieve real-time fleet reconfiguration in order to reach time-changing combat environments and constantly update their techniques. Previous research reveals that lifecycle costs, especially acquisition costs, shrink significantly as a result of fleet modularization. In addition, military field demands and enemy attacks are highly unpredictable and uncertain. Hence, it is of interest to the US Army to investigate the robustness and adaptability of a modular fleet operation system against demand uncertainty. We model the fleet operation management in a stochastic state space model while considering time delays from operational actions, as well as use model predictive control (MPC) to attain real-time optimal operation actions based on the received demands and predicted system status. Analyses on the robustness and adaptability of how a modular vehicle fleet reacts to the demand disturbance and noise have been very limited, although research on operation management and model prediction control have been ongoing for many years. In our current study, we model all the main processes in a fleets operation into an integrated system. These processes include module resupply, vehicle transportation, and on-base assembly, disassembly, reconfiguration (ADR) actions. We also consider the fact that delayed field demands trigger additional demands, which might cause system instability under improper operational strategies. We have designed a predictive control approach that includes an optimizer and a simulation process to monitor and control the fleet operation. Under the identical mission demands and fleet configuration settings, a modular vehicle fleet shows a faster reaction speed than a conventional fleet once demand disturbance is injected. Although our study is inspired by a military application, it is not hard to notice that our system also represents a simplified supply chain structure. Thus, our methodology can also be generalized for civilian applications.
Session: 19-1 Energy Storage and Wind Energy Systems
Thursday, October 12, 01:30 PM - 03:30 PM
DSCC2017-5053 A Real-Time Pseudo-2D Bi-Domain Model of PEM Fuel Cells for Automotive Applications
Mr. Alireza Goshtasbi, Graduater Student Research Assistant, Univ. of Michigan
Dr. Benjamin L. Pence Research Engineer, Ford Motor Company
Dr. Tulga Ersal, Assistant Research Scientist, Univ. of Michigan
(not ARC funded)
Abstract: With the goal of on-line diagnosis for automotive applications in mind, a real-time model of polymer electrolyte membrane (PEM) fuel cell is developed. The model draws from the authors’ previous modeling effort in this area and extends its domain to incorporate transport under the lands. Transport in the catalyst and micro-porous layers, which were previously omitted, are also included in the model. Membrane water transport model is modified accordingly. Moreover, a recently developed homogeneous catalyst layer model is used to describe local oxygen transport resistance in the cathode catalyst layer. Computational efficiency is achieved through spatio-temporal decoupling of the problem, which simplifies the handling of the nonlinear terms. This computational efficiency is demonstrated by a set of simulations that resemble operation under conditions encountered in automotive applications. Moreover, simulation results of the model are in qualitative agreement with earlier computationally intensive modeling studies as well as experimental observations. The current modeling study demonstrates a significant potential for using relatively high-fidelity physics-based models on-line to improve fuel cell performance and durability, which can have a profound impact on its commercialization.
1-1 Mechatronics I
Thursday, October 12, 04:00 PM - 06:00 PM
DSCC2017-5164 Modeling Eddy-Current Damping Force in Magnetic Levitation Systems With Conductors
Mr. Mohammad Khodabakhsh, PhD student, Univ. Of Michigan
Mr. Mehran Ebrahimian, Graduate student, Univ. of Pennsylvania
Prof. Bogdan Epureanu, Assistant Professor, Univ. Of Michigan
(not ARC funded)
Abstract: An analytical method, a model is used to develop a model to calculate steady-state eddy-current damping effects in two configurations of magnetic levitation (maglev) systems. The eddy-current based force (eddy-current force) is used for high precision positioning of a levitated permanent magnet in maglev systems. In these systems, the motion of the levitated permanent magnet and changes of the coil’s currents, generate eddy current in the conductors. The proposed analytical model is used to calculate both effects. A conductive cylindrical shell around the levitated object is implemented as a new technique to generate eddy currents in maglev systems. The model is also employed to obtain eddy-current damping effects in a system with a conductive plate beneath the levitated object. The analytical models match results from high fidelity finite element analysis (FEA) with acceptable accuracy in a wide range of operations. Advantages of the two configurations are discussed.
13-1 Hybrid Electric Vehicles
Thursday, October 12, 04:00 PM - 06:00 PM
DSCC2017-5200 Control and Design Optimization of a Novel Hybrid Electric Powertrain System
Mr. Chenyu Yi, Student, Univ. of Michigan, Ann Arbor
Prof. Bogdan Epureanu, Assistant Professor, Univ. Of Michigan
Abstract: Control and design optimization of hybrid electric powertrains is necessary to maximize the benefits of novel architectures. Previous studies have proposed multiple optimal and nearoptimal control methods, approaches for design optimization, and ways to solve coupled design and control optimization problems for hybrid electric powertrains. This study presents control and design optimization of a novel hybrid electric powertrain architecture to evaluate its performance and potential using physics-based models for the electric machines, the battery and a near-optimal control, namely the equivalent consumption minimization strategy. Design optimization in this paper refers to optimizing the sizes of the powertrain components, i.e. electric machines, battery and final drive. The control and design optimization problem is formulated using nested approach with sequential quadratic programming as design optimization method. Metamodeling is applied to abstract the near-optimal powertrain control model to reduce the computational cost. Fuel economy, sizes of components, and consistency of city and highway fuel economy are reported to evaluate the performance of the powertrain designs. The results suggest an optimal powertrain design and control that grants good performance. The optimal design is shown to be robust and non-sensitive to slight component size changes when evaluated for the near-optimal control.
Session: 16-1 Unmanned, Ground and Surface Robotics I
Friday, October 13, 10:00 AM - 12:00 PM
DSCC2017-5086 A Driver Model for Predicting Human Steering Performance in Teleoperated Path Following of Unmanned Ground Vehicles
Dr. Hossein Mirinejad, Research Fellow, Univ. of Michigan
Dr. Paramsothy Jayakumar, Senior Research Scientist, US Army TARDEC
Dr. Tulga Ersal, Assistant Research Scientist, Univ. of Michigan
Abstract: This paper presents a steering model for predicting human performance in teleoperating unmanned ground vehicles (UGVs). The task of path following, including lane keeping and curve negotiation, is considered for a UGV teleoperation system. Human steering performance in teleoperation is notably different than steering performance in on-board driving conditions due to considerable communication delays in remote teleoperation systems and limited information teleoperators receive from the vehicle sensory system. This paper adopts a cognitive model that was originally developed for a typical highway driving scenario when driver is on board and develops a tuning strategy to adjust the model parameters without human data to reflect the effect of various latencies and UGV speeds on driver performance in a teleoperated path following task. It is shown that the proposed model with tuning strategy i) can adequately capture the trend of changes in driver performance for different teleoperated driving scenarios ii) is able to predict an expert human teleoperator’s performance across different speeds and latencies considered. Thus, the tuned model can be an appropriate candidate to be used in place of human drivers for the simulation-based evaluation of UGV mobility in teleoperation systems.