Jemin George

ClariSense: Clarifying sensor anomalies using social network feeds

The explosive growth in social networks that publish real-time content begs the question of whether their feeds can complement traditional sensors to achieve augmented sensing capabilities. One such capability is to explain anomalous sensor readings. Towards that end, in this paper, we build an automated anomaly clarification service, called ClariSense. It explains sensor anomalies using social network feeds. Explanation goes beyond detection. When a sensor network detects anomalous conditions, our system automatically suggests hypotheses that explain the likely causes of the anomaly to a human by identifying unusual social network feeds that seem to be correlated with the sensor anomaly in time and in space. To evaluate this service, we use real-time data feeds from the California traffic system that shares vehicle count and traffic speed on major California highways at 5 minute intervals. When anomalies are detected, our system automatically diagnoses their root cause by correlating the anomaly with feeds on Twitter. The identified cause is then compared to official traffic and incident reports, showing a great correspondence with ground truth.

Robust Kalman-Bucy Filter

Development of a robust estimator for uncertain stochastic systems under persistent excitation is presented. The given continuous-time stochastic formulation assumes norm bounded parametric uncertainties and excitations. When there are no system uncertainties, the performance of the proposed robust estimator is similar to that of the Kalman-Bucy filter and the proposed approach asymptotically recovers the desired optimal performance in the presence of uncertainties and or persistent excitation.

On localizing urban events with Instagram

This paper develops an algorithm that exploits picture-oriented social networks to localize urban events. We choose picture-oriented networks because taking a picture requires physical proximity, thereby revealing the location of the photographed event. Furthermore, most modern cell phones are equipped with GPS, making picture location, and time metadata commonly available. We consider Instagram as the social network of choice and limit ourselves to urban events (noting that the majority of the world population lives in cities). The paper introduces a new adaptive localization algorithm that does not require the user to specify manually tunable parameters. We evaluate the performance of our algorithm for various real-world datasets, comparing it against a few baseline methods. The results show that our method achieves the best recall, the fewest false positives, and the lowest average error in localizing urban events.

On quality of event localization from social network feeds

Social networks, such as Twitter, carry important information on ongoing events and as such can be viewed as networks of sensors that monitor and report events in the physical world. In this paper, we concern ourselves with the challenge of event localization from Twitter feeds. We explore the quality of information that can be derived either directly or indirectly from microblog entries regarding locations of ongoing events. Contrary to prior work that used Twitter to map regions of large-footprint events, or derived coarse-grained location information, in this paper, we are interested in point-events, such as building fires or car accidents, and aim to pin-point them down to a street address. An algorithm is presented that identifies distinct event signatures in the blogosphere, clusters microblogs based on events they describe, and analyzes the resulting clusters for fine-grained location indicators. An exact event location is then derived by fusing these indicators. To evaluate the quality of derived location information, we use road-traffic-related Twitter feeds from 3 major cities in California and compare automatic event localization within our service to manually obtained ground truth data. Results show a great correspondence between our automatically determined locations and ground-truth.

Joint Localization of Events and Sources in Social Networks

Recent sensor network literature investigated the use of social networks as sensor networks, and formulated a physical event localization problem from social network data. This paper improves on the above results by formulating a joint localization problem of events and sources, leveraging the fact that sources on social networks often have a location affinity: They tend to comment more on events in their locations of interest. While social networks, such as Twitter, do not offer source location information for the majority of sources, we show that our algorithms for jointly inferring source and event location significantly improve localization quality by mutually enhancing location estimation of both events and sources. We evaluate the performance of our algorithm both in simulation and using Twitter data about current events. The results show that joint inference of source and event location allows us to localize many more of the events identified in real-world datasets.

Adaptive disturbance accommodating controller for uncertain stochastic systems

This paper presents a Kalman filter-based adaptive disturbance-accommodating stochastic control scheme for linear uncertain systems to minimize the adverse effects of both model uncertainties and external disturbances. A rigorous stochastic stability analysis reveals a lower bound requirement on system process noise covariance to ensure the stability of the controlled system when the nominal control action on the true plant is unstable. Finally, an adaptive law is synthesized for the selection of stabilizing system process noise covariance. Simulation results are presented where the proposed control scheme is implemented on a two degree-of-freedom helicopter.

Fusion solution for soldier wearable gunfire detection systems

Currently existing acoustic based Gunfire Detection Systems (GDS) such as soldier wearable, vehicle mounted, and fixed site devices provide enemy detection and localization capabilities to the user. However, the solution to the problem of portability versus performance tradeoff remains elusive. The Data Fusion Module (DFM), described herein, is a sensor/platform agnostic software supplemental tool that addresses this tradeoff problem by leveraging existing soldier networks to enhance GDS performance across a Tactical Combat Unit (TCU). The DFM software enhances performance by leveraging all available acoustic GDS information across the TCU synergistically to calculate highly accurate solutions more consistently than any individual GDS in the TCU. The networked sensor architecture provides additional capabilities addressing the multiple shooter/fire-fight problems in addition to sniper detection/localization. The addition of the fusion solution to the overall Size, Weight and Power & Cost (SWaP&C) is zero to negligible. At the end of the first-year effort, the DFM integrated sensor networks performance was impressive showing improvements upwards of 50% in comparison to a single sensor solution. Further improvements are expected when the networked sensor architecture created in this effort is fully exploited.

A mobile self synchronizing smart sensor array for detection and localization of impulsive threat sources

Smart sensors are becoming an integral part of the evolving technology landscape; their ability to share reduced data over networks enables live data fusion, which significantly improves sensor performance and situational awareness. A lightweight, mobile acoustic sensor network has been used as an infrastructure to layer multi-sensor fusion algorithms, for detection of impulsive events such as gunfire or explosions. The system can create actionable information within seconds, and can be used to direct assets such as unmanned aerial vehicles (UAVs) to specific coordinates, for eyes-on assessment in under a minute. The sensor array will be discussed in terms of its three primary components: the smart sensors, the synchronization network, and the fusion algorithms. Performance of the array from recent tests will be examined with respect to small arms and simulated mortar fire, and producing actionable information. In addition, test results will be discussed in context of autonomous control of UAV assets and potential applications.

Binary consensus through binary communication

The problem of binary consensus for an undirected, synchronous, fixed topology network through noiseless, binary communication is the focus of this paper. We propose a binary consensus protocol that guarantees network convergence to the initial network majority via binary communication among one-hop neighbors. The proposed protocol requires the nodes to keep track of an internal state, which accounts for the local disagreement and the local average of the network vote. Convergence of the algorithm to the initial global majority is proved using the lossy S-procedure. Numerical simulations are included to demonstrate the performance of the proposed protocol.

Spacecraft attitude estimation using adaptive gaussian sum filter

This paper is concerned with improving the attitude estimation accuracy by implementing an adaptive Gaussian sum filter where the a posteriori density function is approximated by a sum of Gaussian density functions. Compared to the traditional Gaussian sum filter, this adaptive approach utilizes the Fokker-Planck-Kolmogorov residual minimization to update the weights associated with different components of the Gaussian mixture model. Updating the weights provides an accurate approximation of the a posteriori density function and thus superior estimates. Simulation results show that updating the weights during the propagation stage not only provides better estimates between the observations but also provides superior estimator performance where the measurements are ambiguous.