Amit Surana

A sensor-utility-network method for estimation of occupancy in buildings

We introduce the sensor-utility-network (SUN) method for occupancy estimation in buildings. Based on inputs from a variety of sensor measurements, along with historical data regarding building utilization, the SUN estimator produces occupancy estimates through the solution of a receding-horizon convex optimization problem. State-of-the-art on-line occupancy algorithms rely on indirect measurements, such as CO2 levels, or people counting sensors which are subject to significant errors and cost. The newly proposed method was evaluated via experiments in an office building environment. Estimation accuracy is shown to improve significantly when all available data is incorporated in the estimator. In particular, it is found that the average estimation error at the building level is reduced from 70% to 11% using the SUN estimator, when compared to the naive approach that relies solely on flow measurements.

Control, estimation and optimization of energy efficient buildings

Commercial buildings are responsible for a significant fraction of the energy consumption and greenhouse gas emissions in the U.S. and worldwide. Consequently, the design, optimization and control of energy efficient buildings can have a tremendous impact on energy cost and greenhouse gas emission. Buildings are complex, multi-scale in time and space, multi-physics and highly uncertain dynamic systems with wide varieties of disturbances. Recent results have shown that by considering the whole building as an integrated system and applying modern estimation and control techniques to this system, one can achieve greater efficiencies than obtained by optimizing individual building components such as lighting and HVAC. We consider estimation and control for a distributed parameter model of a multi-room building. In particular, we show that distributed parameter control theory, coupled with high performance computing, can provide insight and computational algorithms for the optimal placement of sensors and actuators to maximize observability and controllability. Numerical examples are provided to illustrate the approach. We also discuss the problems of design and optimization (for energy and CO2 reduction) and control (both local and supervisory) of whole buildings and demonstrate how sensitivities can be used to address these problems.

Koopman operator based observer synthesis for control-affine nonlinear systems

We propose a new observer form based on Koopman operator theoretic framework for input-output nonlinear systems with control affine inputs. Based on this observer form, we describe an observer synthesis framework which exploits estimation techniques developed for Lipschitz systems and bilinear systems. We also formulate nonlinear observability rank condition in terms of the Koopman observer form, and numerically illustrate the benefits of the proposed framework.

Reduced order modeling for contaminant transport and mixing in building systems: A case study using dynamical systems techniques

In this paper we propose a Lagrangian coherent structures (LCS) based approach to modeling and estimation of contaminant transport and mixing in large indoor spaces in buildings. Specifically, we show how the knowledge of LCS can be exploited to enhance proper orthogonal decomposition (POD) based model reduction, sensor placement and comparing effect of different control schemes. We illustrate this approach in a three-dimensional room equipped with a mechanical ventilation system.

Anomaly detection in videos: A dynamical systems approach

We demonstrate a dynamical system framework based on motion patterns for detecting anomalous individual and group behavior in complex videos. We first describe a framework based on trajectory modeling, in which coarse statistical models are used to capture global motion patterns, and are employed in change detection to identify anomalous behavior at the object level. Our multi-target tracking framework combines geometric active contours with particle filtering to effectively deal with occlusions and clutter in the environment. In crowded scenes, however, such object level representation can become extremely unreliable: to deal with this we instead use of low-level motion features (e.g., optical flow) to capture group behavior. To keep the problem tractable, we utilize a subspace system identification method based on the Hankel matrix to extract relevant low order dynamics of these noisy features. The spectral properties of the Hankel matrix encode useful information about the dynamics, and can detect anomalous group behavior. In order to efficiently compute these spectral properties, we employ a randomized algorithm for singular value decomposition. Both approaches are demonstrated to robustly detect anomalous behavior in realistic indoor and outdoor videos.

Experimental Implementation of Spectral Multiscale Coverage and Search Algorithms for Autonomous UAVs

We discuss the implementation of Spectral Multiscale Coverage (SMC) based multi-vehicle control and coordination for coverage and search missions by autonomous UAVs. The SMC algorithm gives rise to multiscale vehicle trajectories leading to efficient coverage of a given area and thereby making it useful for search algorithms that are robust to sensor errors and terrain uncertainties. We provide a functional summary of the SMC framework and address its practical implementation. The practical feasibility of the SMC approach is demonstrated via several coverage problems using high fidelity software-in-the-loop (SIL) simulations and experimental flight tests conducted using an electric helicopter.

Scalable uncertainty quantification in complex dynamic networks

In this paper we address the problem of uncertainty management for robust design, and verification of large dynamic networks whose performance is affected by an equally large number of uncertain parameters. Many such networks (e.g. power, thermal and communication networks) are often composed of weakly interacting subnetworks. We propose an iterative scheme that exploits such weak interconnections to overcome dimensionality curse associated with traditional uncertainty quantification methods (e.g. Quasi Monte Carlo, Probabilistic Collocation) and accelerate uncertainty propagation in systems with large number of uncertain parameters. This approach relies on integrating graph theoretic methods and waveform relaxation with traditional uncertainty quantification techniques like probabilistic collocation and polynomial chaos. We analyze convergence properties of this scheme and illustrate it on two examples.

Unsupervised inverse reinforcement learning with noisy data

In this paper we propose an approach for unsupervised Inverse Reinforcement Learning (IRL) with noisy data using a hidden variable Markov Decision Processes (hMDP) representation. hMDP accounts for observation uncertainty by using a hidden state variable. We develop a nonparametric Bayesian IRL technique for hMDP based on Dirichlet Processes mixture model. We provide an efficient Markov Chain Monte Carlo based sampling algorithm whereby one can automatically cluster noisy data into different behaviors, and estimate the underlying reward parameters per cluster. We demonstrate our approach for unsupervised learning, and prediction and classification of agent behaviors in a simulated surveillance scenario.

A static coverage algorithm for locational optimization

We propose multiscale metrics to capture the quality of coverage by a static configuration of agents. This metric is used for the locational optimization of sensor networks. Agent configurations that minimize the multiscale coverage metric are an alternative to the well-known centroidal voronoi tesselations. Other applications include quantization and clustering analysis. We demonstrate the performance of the algorithm on various examples.