Real-Time and Embedded Systems Lab (mLAB)Copyright (c) 2017 University of Pennsylvania All rights reserved.
http://repository.upenn.edu/mlab_papers
Recent documents in Real-Time and Embedded Systems Lab (mLAB)en-usSat, 10 Jun 2017 01:47:12 PDT3600Smooth Operator: Control using the Smooth Robustness of Temporal Logic
http://repository.upenn.edu/mlab_papers/100
http://repository.upenn.edu/mlab_papers/100Thu, 08 Jun 2017 13:29:51 PDT
Modern control systems, like controllers for swarms of quadrotors, must satisfy complex control objectives while withstanding a wide range of disturbances, from bugs in their software to attacks on their sensors and changes in their environments. These requirements go beyond stability and tracking, and involve temporal and sequencing constraints on system response to various events. This work formalizes the requirements as formulas in Metric Temporal Logic (MTL), and designs a controller that maximizes the robustness of the MTL formula. Formally, if the system satisfies the formula with robustness r, then any disturbance of size less than r cannot cause it to violate the formula. Because robustness is not differentiable, this work provides arbitrarily precise, infinitely differentiable, approximations of it, thus enabling the use of powerful gradient descent optimizers. Experiments on a temperature control example and a two-quadrotor system demonstrate that this approach to controller design outper- forms existing approaches to robustness maximization based on Mixed Integer Linear Programming and stochastic heuristics. Moreover, it is not constrained to linear systems.
]]>
Yash Vardhan Pant et al.CPS TheoryCPS Formal MethodsCPS Model-Based DesignComputer-Aided Design for Safe Autonomous Vehicles
http://repository.upenn.edu/mlab_papers/99
http://repository.upenn.edu/mlab_papers/99Mon, 01 May 2017 20:59:35 PDTMatthew O'Kelly et al.CPS AutoCPS Model-Based DesignTechnical Report: Control Using the Smooth Robustness of Temporal Logic
http://repository.upenn.edu/mlab_papers/98
http://repository.upenn.edu/mlab_papers/98Wed, 01 Mar 2017 09:06:13 PST
Cyber-Physical Systems must withstand a wide range of errors, from bugs in their software to attacks on their physical sensors. Given a formal specification of their desired behavior in Metric Temporal Logic (MTL), the robust semantics of the specification provides a notion of system robustness that can be calculated directly on the output behavior of the system, without explicit reference to the various sources or models of the errors. The robustness of the MTL specification has been used both to verify the system offline (via robustness minimization) and to control the system online (to maximize its robustness over some horizon). Unfortunately, the robustness objective function is difficult to work with: it is recursively defined, non-convex and non-differentiable. In this paper, we propose smooth approximations of the robustness. Such approximations are differentiable, thus enabling us to use powerful off-the- shelf gradient descent algorithms for optimizing it. By using them we can also offer guarantees on the performance of the optimization in terms of convergence to minima. We show that the approximation error is bounded to any desired level, and that the approximation can be tuned to the specification. We demonstrate the use of the smooth robustness to control two quad-rotors in an autonomous air traffic control scenario, and for temperature control of a building for comfort.
]]>
Yash Vardhan Pant et al.CPS TheoryRelaxed decidability and the robust semantics of Metric Temporal Logic
http://repository.upenn.edu/mlab_papers/97
http://repository.upenn.edu/mlab_papers/97Thu, 23 Feb 2017 08:46:32 PST
Relaxed notions of decidability widen the scope of automatic verification of hybrid systems. In quasi-decidability and $\delta$-decidability, the fundamental compromise is that if we are willing to accept a slight error in the algorithm's answer, or a slight restriction on the class of problems we verify, then it is possible to obtain practically useful answers. This paper explores the connections between relaxed decidability and the robust semantics of Metric Temporal Logic formulas. It establishes a formal equivalence between the robustness degree of MTL specifications, and the imprecision parameter $\delta$ used in $\delta$-decidability when it is used to verify MTL properties. We present an application of this result in the form of an algorithm that generates new constraints to the $\delta$-decision procedure from falsification runs, which speeds up the verification run. We then establish new conditions under which robust testing, based on the robust semantics of MTL, is in fact a quasi-semidecision procedure. These results allow us to delimit what is possible with fast, robustness-based methods, accelerate (near-)exhaustive verification, and further bridge the gap between verification and simulation.
]]>
Houssam Abbas et al.CPS Model-Based DesignCPS Formal MethodsCPS TheoryData Predictive Control for building energy management
http://repository.upenn.edu/mlab_papers/96
http://repository.upenn.edu/mlab_papers/96Fri, 27 Jan 2017 22:16:34 PST
Decisions on how to best optimize energy systems operations are becoming ever so complex and conflicting, that model-based predictive control (MPC) algorithms must play an important role. However, a key factor prohibiting the widespread adoption of MPC in buildings, is the cost, time, and effort associated with learning first-principles based dynamical models of the underlying physical system. This paper introduces an alternative approach for implementing finite-time receding horizon control using control-oriented data-driven models. We call this approach Data Predictive Control (DPC). Specifically, by utilizing separation of variables, two novel algorithms for implementing DPC using a single regression tree and with regression trees ensembles (random forest) are presented. The data predictive controller enables the building operator to trade off energy consumption against thermal comfort without having to learn white/grey box models of the systems dynamics. We present a comprehensive numerical study which compares the performance of DPC with an MPC based energy management strategy, using a single zone building model. Our results demonstrate that performance of DPC is comparable to an MPC controller, with only 3.8% additional cost in terms of optimal objective function and within 95% in terms of R2 score, thereby making it an alluring alternative to MPC, whenever the associated cost of learning the model is high.
]]>
Achin Jain et al.CPS Efficient BuildingsCPS Real-TimeHigh-Level Modeling for Computer-Aided Clinical Trials of Medical Devices
http://repository.upenn.edu/mlab_papers/95
http://repository.upenn.edu/mlab_papers/95Sun, 27 Nov 2016 14:34:23 PSTHoussam Abbas et al.CPS MedicalCPS Model-Based DesignRobust Model Predictive Control for Non-Linear Systems with Input and State Constraints Via Feedback Linearization
http://repository.upenn.edu/mlab_papers/94
http://repository.upenn.edu/mlab_papers/94Thu, 17 Nov 2016 12:14:14 PST
Robust predictive control of non-linear systems under state estimation errors and input and state constraints is a challenging problem, and solutions to it have generally involved solving computationally hard non-linear optimizations. Feedback linearization has reduced the computational burden, but has not yet been solved for robust model predictive control under estimation errors and constraints. In this paper, we solve this problem of robust control of a non-linear system under bounded state estimation errors and input and state constraints using feedback linearization. We do so by developing robust constraints on the feedback linearized system such that the non-linear system respects its constraints. These constraints are computed at run-time using online reachability, and are linear in the optimization variables, resulting in a Quadratic Program with linear constraints. We also provide robust feasibility, recursive feasibility and stability results for our control algorithm. We evaluate our approach on two systems to show its applicability and performance.
]]>
Yash Vardhan Pant et al.CPS TheoryComputer Aided Clinical Trials for Implantable Cardiac Devices
http://repository.upenn.edu/mlab_papers/93
http://repository.upenn.edu/mlab_papers/93Wed, 16 Nov 2016 09:02:17 PST
In this effort we investigate the design and use of physiological and device models to conduct pre-clinical trials to provide early insight in the design and execution of the actual clinical trial. Computer models of physiological phenomena like cardiac electrical activity can be extremely complex. However, when the purpose of the model is to interact with a medical device, then it becomes sufficient to model the measurements that the device makes, e.g. the intra-cardiac electrograms (EGMs) that an Implantable Cardioverter Defibrillator (ICD) measures. We present a probabilistic generative model of EGMs, capable of generating exemplars of various arrhythmias. The model uses deformable shape templates, or motifs, to capture the variability in EGM shapes within one EGM channel, and a cycle length parameter to capture the variability in cycle length in one EGM channel. The relation between EGM channels, which is essential for determining whether the current arrhythmia is potentially fatal, is captured by a time-delayed Markov chain, whose states model the various combinations of (learned) motifs. The heart model is minimally parameterized and is learned from real patient data. Thus the statistics of key features reflect the statistics of a real cohort, but the model can also generate rare cases and new combinations from the inferred probabilities. On the device end, algorithms for signal sensing, detection and discrimination for major ICD manufacturers have been implemented both in simulation and on hardware platforms. The generated arrhythmia episodes are used as input to both the modeled ICD algorithms and real ICDs as part of a Computer Aided Clinical Trial (CACT). In a CACT, a computer model simulates the inputs to the device (such as a new, investigational ICD), and the device’s performance is evaluated. By incorporating these results into the appropriate statistical framework, the Computer Aided Clinical Trial results can serve as regulatory evidence when planning and executing an actual clinical trial. We demonstrate this by conducting a mock trial similar to the 2005-2010 RIGHT trial which compared the discrimination algorithms from two major ICD manufacturers. The results of the CACT clearly demonstrate that the failed outcome of the RIGHT trial could have been predicted and provides statistical support for deeper results that could have been captured prior to the trial.
]]>
Houssam Abbas et al.CPS MedicalData Predictive Control for Peak Power Reduction
http://repository.upenn.edu/mlab_papers/92
http://repository.upenn.edu/mlab_papers/92Sat, 17 Sep 2016 11:16:37 PDT
Decisions on how best to optimize today's energy systems operations are becoming ever so complex and conflicting such that model-based predictive control algorithms must play a key role. However, learning dynamical models of energy consuming systems such as buildings, using grey/white box approaches is very cost and time prohibitive due to its complexity. This paper presents data-driven methods for making control-oriented model for peak power reduction in buildings. Specifically, a data predictive control with regression trees (DPCRT) algorithm, is presented. DPCRT is a finite receding horizon method, using which the building operator can optimally trade off peak power reduction against thermal comfort without having to learn white/grey box models of the systems dynamics. We evaluate the performance of our method using a DoE commercial reference virtual test-bed and show how it can be used for learning predictive models with 90% accuracy, and for achieving 8.6% reduction in peak power and costs.
]]>
Achin Jain et al.CPS Efficient BuildingsCPS Real-TimeCo-Design of Anytime Computation and Robust Control
http://repository.upenn.edu/mlab_papers/91
http://repository.upenn.edu/mlab_papers/91Thu, 05 May 2016 09:05:50 PDTYash Vardhan Pant et al.CPS Real-TimeCPS Embedded ControlCPS TheoryBenchmark: Nonlinear Hybrid Automata Model of Excitable Cardiac Tissue
http://repository.upenn.edu/mlab_papers/90
http://repository.upenn.edu/mlab_papers/90Wed, 23 Mar 2016 11:46:19 PDT
Implantable cardiac devices like pacemakers and defibrillators are life-saving medical devices. To verify their functionality, there is a need for heart models that can simulate interesting phenomena and are relatively computationally tractable. In this benchmark we implement a model of the electrical activity in excitable cardiac tissue as a network of nonlinear hybrid automata. The model has previously been shown to simulate fast arrhythmias. The hybrid automata are arranged in a square n-by-n grid and communicate via their voltages. Our Matlab implementation allows the user to specify any size of model $n$, thus rendering it ideal for benchmarking purposes since we can study tool efficiency as a function of size. We expect the model to be used to analyze parameter ranges and network connectivity that lead to dangerous heart conditions. It can also be connected to device models for device verification.
]]>
Houssam Abbas et al.CPS MedicalAutomated Closed-Loop Model Checking of Implantable Pacemakers using Abstraction Trees
http://repository.upenn.edu/mlab_papers/89
http://repository.upenn.edu/mlab_papers/89Tue, 22 Mar 2016 11:09:32 PDT
Autonomous medical devices such as implantable cardiac pacemakers are capable of diagnosing the patient condition and delivering therapy without human intervention. Their ability to autonomously affect the physiological state of the patient makes them safety-critical. Sufficient evidence for the safety and efficacy of the device software, which makes these autonomous decisions, should be provided before these devices can be released on the market. Formal methods like model checking can provide safety evidence that the devices can safely operate under a large variety of physiological conditions. The challenge is to develop physiological models that are general enough to cover the large variability of human physiology, and also expressive enough to provide physiological contexts to counter-examples returned by the model checker. In this paper, the authors develop a set of physiological abstraction rules that introduce physiological constraints to heart models. By applying these abstraction rules to a initial set of heart models, an abstraction tree is created. The root model covers all possible inputs to a pacemaker and derived models cover inputs from different heart conditions. If a counter-example is returned by the model checker, the abstraction tree is traversed so that the most concrete counter-example(s) with physiological contexts can be returned to the domain experts for validity check. The abstraction tree framework replaces the manual abstraction and refinement framework, which reduced the amount of domain knowledge required to perform closed-loop model checking. It encourages the use of model checking during the development of autonomous medical devices, and identifies safety risks earlier in the design process.
]]>
Zhihao Jiang et al.CPS MedicalTech Report: Robust Model Predictive Control for Non-Linear Systems with Input and State Constraints Via Feedback Linearization
http://repository.upenn.edu/mlab_papers/88
http://repository.upenn.edu/mlab_papers/88Tue, 15 Mar 2016 19:39:14 PDT
Robust predictive control of non-linear systems under state estimation errors and input and state constraints is a challenging problem, and solutions to it have generally involved solving computationally hard non-linear optimizations. Feedback linearization has reduced the computational burden, but has not yet been solved for robust model predictive control under estimation errors and constraints. In this paper, we solve this problem of robust control of a non-linear system under bounded state estimation errors and input and state constraints using feedback linearization. We do so by developing robust constraints on the feedback linearized system such that the non-linear system respects its constraints. These constraints are computed at run-time using online reachability, and are linear in the optimization variables, resulting in a Quadratic Program with linear constraints. We also provide robust feasibility, recursive feasibility and stability results for our control algorithm. We evaluate our approach on two systems to show its applicability and performance
]]>
Yash Vardhan Pant et al.CPS TheoryTowards Model Checking of Implantable Cardioverter Defibrillators
http://repository.upenn.edu/mlab_papers/87
http://repository.upenn.edu/mlab_papers/87Thu, 03 Mar 2016 13:54:16 PST
Ventricular Fibrillation is a disorganized electrical excitation of the heart that results in inadequate blood flow to the body. It usually ends in death within a minute. A common way to treat the symptoms of fibrillation is to implant a medical device, known as an Implantable Cardioverter Defibrillator (ICD), in the patient's body. Model-based verification can supply rigorous proofs of safety and efficacy. In this paper, we build a hybrid system model of the human heart+ICD closed loop, and show it to be a STORMED system, a class of o-minimal hybrid systems that admit finite bisimulations. In general, it may not be possible to compute the bisimulation. We show that approximate reachability can yield a finite simulation for STORMED systems, and that certain compositions respect the STORMED property. The results of this paper are theoretical and motivate the creation of concrete model checking procedures for STORMED systems.
]]>
Houssam Abbas et al.CPS MedicalCPS Formal MethodsCloudMat: Context-aware Personalization of Fitness Content
http://repository.upenn.edu/mlab_papers/86
http://repository.upenn.edu/mlab_papers/86Tue, 26 Jan 2016 05:21:27 PST
Digital video content via broadcast television, Internet and other content distribution networks provide limited interaction for fitness and wellness activities. The content delivery is one-way only and provides no personalization to the pace, programming and progress of the user’s exercise routine. Furthermore, the content is to be viewed only on a screen which makes it awkward and incompatible with full-body activities such as yoga, pilates and T’ai chi. We present CloudMat, a system for context-aware personalization of fitness content with cloudenabled connected surfaces. CloudMat provides real-time closedloop feedback between the state of the user on the physical mat and the state of the content in the cloud service. Content is tagged with actuation signals where events are delegated from the screen to display on an electroluminescent lighting layer on the mat, which provides spatial guidance to the end-user. Through the sensor-layer embedded in the mat, the physical interface captures the pose and timing of the user activity and relays it to the Context-aware Personalization cloud service. This service coordinates sensing and actuation between the content stream and mat by generating pose templates and metadata files about the exercise routine to be delivered to the user. Through this interactive process between the physical mat and the content service, the feedback provided by the user performing the routine continuously adapts the pace and programming to maintain the desired user experience. We demonstrate the utility of the system and evaluate the system performance with a case study on interactive yoga.
]]>
Kuk Jin Jang et al.CPS Internet of ThingsCampus-Wide Integrated Building Energy Simulation
http://repository.upenn.edu/mlab_papers/85
http://repository.upenn.edu/mlab_papers/85Fri, 15 Jan 2016 13:23:02 PST
Effective energy management for large campus facilities is becoming increasingly complex as modern heating and cooling systems comprise of several hundred subsystems interconnected to each other. Building energy simulators like EnergyPlus are exceedingly good at modeling a single building equipped with a standalone HVAC equipment. However, the ability to simulate a large campus and to control the dynamics and interactions of the subsystems is limited or missing altogether. In this paper, we use the Matlab-EnergyPlus MLE+ tool we developed, to extend the capability of EnergyPlus to co-simulate a campus with multiple buildings connected to a chilled water distribution to a central chiller plant with control systems in Matlab. We present the details of how this simulation can be set-up and implemented using MLE+'s Matlab/Simulink block. We utilize the virtual campus test-bed to evaluate the performance of several demand response strategies. We also describe a coordinated demand response scheme which can lead to load curtailment during a demand response event while minimizing thermal discomfort.
]]>
Willy Bernal et al.CPS Efficient BuildingsAPEX: Autonomous Vehicle Plan Verification and Execution
http://repository.upenn.edu/mlab_papers/84
http://repository.upenn.edu/mlab_papers/84Fri, 15 Jan 2016 13:18:47 PST
Autonomous vehicles (AVs) have already driven millions of miles on public roads, but even the simplest scenarios have not been certified for safety. Current methodologies for the verification of AV's decision and control systems attempt to divorce the lower level, short-term trajectory planning and trajectory tracking functions from the behavioral rules-based framework that governs mid-term actions. Such analysis is typically predicated on the discretization of the state space and has several limitations. First, it requires that a conservative buffer be added around obstacles such that many feasible plans are classified as unsafe. Second, the discretized controllers modeled in this analysis require several refinement steps before being implementable on an actual AV, and typically do not allow the specification of comfort-related properties on the trajectories. In contrast, consumer-ready AVs use motion planning algorithms that generate smooth trajectories. While viable algorithms exist for the generation of smooth trajectories originating from a single state, analysis should consider that the AV faces state estimation errors and disturbances. Third, verification is restricted to a discretized state space with fixed-size cells; this assumption can artificially limit the set of available trajectories if the discretization is too coarse. Conversely, too fine of a discretization renders the problem intractable for automated analysis. This work presents a new verification tool, APEX, which investigates the combined action of a behavioral planner and state lattice-based motion planner to guarantee a safe vehicle trajectory is chosen. In APEX, decisions made at the behavioral layer can be traced through to the spatio-temporal evolution of the AV and verified. Thus, there is no need to create abstractions of the AV's controllers, and aggressive trajectories required for evasive maneuvers can be accurately investigated.
]]>
Matthew O'Kelly et al.CPS AutoCPS Formal MethodsDR-Advisor: A Data-Driven Demand Response Recommender System
http://repository.upenn.edu/mlab_papers/83
http://repository.upenn.edu/mlab_papers/83Fri, 15 Jan 2016 13:18:42 PST
Demand response (DR) is becoming increasingly important as the volatility on the grid continues to increase. Current DR ap- proaches are predominantly completely manual and rule-based or involve deriving first principles based models which are ex- tremely cost and time prohibitive to build. We consider the problem of data-driven end-user DR for large buildings which involves predicting the demand response baseline, evaluating fixed rule based DR strategies and synthesizing DR control actions. The challenge is in evaluating and taking control decisions at fast time scales in order to curtail the power consumption of the building, in return for a financial reward. We provide a model based control with regression trees algorithm (mbCRT), which allows us to perform closed-loop control for DR strategy synthesis for large commercial buildings. Our data-driven control synthesis algorithm outperforms rule-based DR by 17% for a large DoE commercial reference building and leads to a curtailment of 380kW and over $45, 000 in savings. Our methods have been integrated into an open source tool called DR-Advisor, which acts as a recommender system for the building’s facilities manager and provides suitable control actions to meet the desired load curtailment while main- taining operations and maximizing the economic reward. DR-Advisor achieves 92.8% to 98.9% prediction accuracy for 8 buildings on Penn’s campus. We compare DR-Advisor with other data driven methods and rank 2nd on ASHRAE’s benchmarking data-set for energy prediction.
]]>
Madhur Behl et al.CPS Efficient BuildingsSometimes, Money Does Grow On Trees: Data-Driven Demand Response with DR-Advisor
http://repository.upenn.edu/mlab_papers/82
http://repository.upenn.edu/mlab_papers/82Fri, 15 Jan 2016 13:07:22 PST
Real-time electricity pricing and demand response has become a clean, reliable and cost-effective way of mitigating peak demand on the electricity grid. We consider the problem of end-user demand response (DR) for large commercial buildings which involves predicting the demand response baseline, evaluating fixed DR strategies and synthesizing DR control actions for load curtailment in return for a financial reward. Using historical data from the building, we build a family of regression trees and learn data-driven models for predicting the power consumption of the building in real-time. We present a method called DR-Advisor called DR-Advisor, which acts as a recommender system for the building's facilities manager and provides suitable control actions to meet the desired load curtailment while maintaining operations and maximizing the economic reward. We evaluate the performance of DR-Advisor for demand response using data from a real office building and a virtual test-bed.
]]>
Madhur Behl et al.CPS Efficient BuildingsCPS Model-Based DesignSometimes, Money Does Grow on Trees: DR-Advisor, A Data Driven Demand Response Recommender System
http://repository.upenn.edu/mlab_papers/81
http://repository.upenn.edu/mlab_papers/81Fri, 15 Jan 2016 13:07:20 PST
Unprecedented amounts of information from millions of smart meters and thermostats installed in recent years has left the door open for better understanding, analyzing and using the insights that data can provide, about the power consumption patterns of a building. The challenge with using data-driven approaches, is to close the loop for near real-time control and decision making in large buildings. Furthermore, providing a technological solution alone is not enough, the solution must also be human centric. We consider the problem of end-user demand response for commercial buildings. Using historical data from the building, we build a family of regression trees based models for predicting the power consumption of the building in real-time. We have built DR-Advisor, a recommender system for the building's facilities manager, which provides optimal control actions to meet the required load curtailment while maintaining building operations and maximizing the economic reward.
]]>
Madhur Behl et al.CPS Efficient BuildingsCPS Internet of Things