Chinya Ravishankar - Academia.edu (original) (raw)

Papers by Chinya Ravishankar

Proceedings of the 25th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems

Consider objects moving in a road network (e.g., groups of people or delivery vehicles), who may ... more Consider objects moving in a road network (e.g., groups of people or delivery vehicles), who may be free to choose routes, yet be required to arrive at certain locations at certain times. Such objects may need to assemble in groups within the network (friends meet while visiting a city, vehicles need to exchange items or information) without violating arrival constraints. Planning for such assemblies is hard when the network or the number of objects is large. Conversely, discovering actual or potential assemblies of such objects is important in many surveillance, security, and lawenforcement applications. This can be hard when object arrival observations are sparse due to inadequate sensor coverage or object countermeasures. We propose the novel class of assembly queries to model these scenarios, and present a unified scheme that addresses both of these complementary challenges. Given a set of objects and arrival constraints, we show how to first obtain the set of all possible locations visited by each moving object (the travel corridor), and then determine all possible assemblies, including the participants, locations, and durations. We present a formal model for various tracking strategies and several algorithms for using these strategies. We achieve excellent performance on these queries by preprocessing the network, using Contraction Hierarchies. Experimental results on real-world road networks show that we can efficiently and rapidly infer assembly information for very large networks and object groups.

Stochastic route planning is a hard problem, since it deals with uncertain edge weights, usually ... more Stochastic route planning is a hard problem, since it deals with uncertain edge weights, usually modeled as probability distributions. Stochastic shortest path queries are very expensive, as they must compute convolutions of edge weight distributions, whose representations can have a major impact on query costs. Effective speedup techniques for shortest path queries exist for deterministic edge weights, but their extensions to stochastic settings have had limited success, and real-time stochastic routing queries remain beyond reach. We introduce the tiering technique for Contraction and Edge Hierarchies (CHs and EHs) to address this challenge. We divide the hierarchy into tiers, and represent edge weights in each tier in ways that permit effective tradeoffs between accuracy, convolution costs, and space use. We show how to use Gaussians to approximate histograms, and bound errors using the KL divergence and Hellinger distance measures. We develop Uncertain Contraction Hierarchies (UCHs) and Uncertain Edge Hierarchies (UEHs) using these methods, and show that they improve both CH and EH performance for three different stochastic query types: probabilistic budget routes, non-dominated routes, and routes to minimize the mean-risk objective. We evaluate our methods using real-world data from Mapbox Traffic Data for a section of Los Angeles. Finally, our results show that query times for EHs can be competitive with CHs for stochastic edge weights, contrary to current belief. CCS CONCEPTS • Theory of computation → Shortest paths; Routing and network design problems.

Electric Vehicle (EV) battery capacity is limited, so EV routing must trade o travel time for ene... more Electric Vehicle (EV) battery capacity is limited, so EV routing must trade o travel time for energy consumption, which grows quadratically with speed. Current multi-parameter EV routing methods assume accurate estimates of time and energy consumed, but current models for obtaining these estimates cannot capture this time-energy tradeo in a suciently exible way. We present a new approach to EV modeling that addresses such shortcomings. Conventional wisdom holds that models operating at ner time granularities yield better energy consumption estimates. We rst show that such is not necessarily the case, by dening a new structuring abstraction for vehicle speed proles called phases, which models energy consumption accurately at lower temporal granularity. We also address the challenge of generating speed proles for planned trips with realistic variance in travel times and energy consumed. Our method combines the phase abstraction with Markov chains and kernel density estimation to learn these variations, and construct realistic vehicle speed proles for real-world routes. Using 52 hours of driving data collected on a Nissan Leaf, we show that our model achieves a per-trip accuracy better than even that of current microscopic models and generates speed proles that accurately model time and energy consumption at the trip level. CCS CONCEPTS • Information systems → Global positioning systems; • Applied computing → Transportation.

A common approach to zmprovzng the performance of statzstzcal query processzng 2s t o use precomp... more A common approach to zmprovzng the performance of statzstzcal query processzng 2s t o use precomputed results Another lower-level approach would be t o redeszgn the storage structure f o r stattstzcal databases. Thzs avenue as relatzvely unexplored. The objectaue of thzs paper zs t o present a physzcal storage structure for statzstzcal databases, whose desagn zs motzvated b y the characterzstzcs of statzstzcal querzes We show that our proposal enhances multz-attrzbute clusterzng eficzency, and zmproves the performance of statzstzcal and aggregatzonal querzes Thzs i ustomzzed structure reduces the amount of 1/0 zncurred durzng statzstzcal query processzng, thus decreaszng Ihc response tzme *'l'his work was supported in part by the Consortium for tnternational Earth Science Information Networking. t'l'he material contained in this paper may be covered by a pending patent application.

Over the past decade, spatial and spatio-temporal databases have found important applications in ... more Over the past decade, spatial and spatio-temporal databases have found important applications in areas such as urban planning, geographical information systems (GIS), computer aided design (CAD) and NASA's earth observation systems. Much work exists on conventional spatial and spatio-temporal queries, but there has been growing interest in evaluating such queries in the presence of movement or uncertainty. First, advances in GPS technologies have created a variety of new applications, such as location-based services, vehicle fleet tracking, and E911 emergency service, which issue complex spatio-temporal queries on the trajectories of moving objects. Second, while it is typically assumed the positional attributes of spatial objects are precisely known, they may be known only approximately in practice, with uncertainty depending on the nature of the measurements and the source of data. The presence of movement or uncertainty requires new techniques to evaluate spatial and spatio-temporal queries. First, unlike traditional spatial databases whose volume is relatively static, the volume of the trajectory data grows rapidly, requiring new and efficient indexing methods to ensure quick responses. Second, these new applications give rise to new types of queries, which are not supported in the traditional spatial databases. Finally, since uncertainty may affect the accuracy of the answer to queries, traditional techniques are inadequate to evaluate spatial data with uncertainty. In this thesis, we present new efficient techniques to process spatial and spatio-temporal queries in the presence of uncertainty and movement. First, we present an efficient spatio-temporal index method based on polynomial approximation for search on trajectory data of moving objects. We show that our technique can process trajectories in both offline and online fashion, with performance superior to previous approaches. Second, we propose point-wise dense region queries, a new type of query, to identify regions with a high concentration of moving objects. We develop two efficient methods to exactly or approximately evaluate point-wise dense region queries. Finally, we propose the notion of probabilistic spatial query to handle uncertainty in spatial databases. We also present a new probabilistic index structure to evaluate probabilistic spatial queries efficiently.

Proceedings of the 26th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, 2018

E ciently querying large trajectory datasets is a challenge of growing importance. Abstracting tr... more E ciently querying large trajectory datasets is a challenge of growing importance. Abstracting trajectory segments with minimum bounding boxes and indexing them in R-Trees results in a high false positive rate due to high dead space. Space lling curves (SFCs), which have excellent locality preserving and dimensionality reduction properties, have been shown to be e ective for indexing points in space. However, they can yield a high false positive count and slow query times if used to index trajectory segments. Our work shows how to use SFCs to index trajectory polylines. In our experiments, the proposed method runs 2-15 times faster than other state-of-the-art approaches.

Deep Blue (University of Michigan), 1992

In this paper, we explore the constraints of a new problem: that of coordinating network troubles... more In this paper, we explore the constraints of a new problem: that of coordinating network troubleshootingamong peer administrative domains or Internet Service Providers, and untrusted observers. Allowing untrustedobservers permits any entity to report problems, whether it is a Network Operations Center (NOC), end-user,or application.Our goals here are to define the inter-domain coordination problem clearly, and to develop an architecturewhich allows

ACM Transactions on Database Systems, Mar 15, 2019

Current Bloom Filters tend to ignore Bayesian priors as well as a great deal of useful informatio... more Current Bloom Filters tend to ignore Bayesian priors as well as a great deal of useful information they hold, compromising the accuracy of their responses. Incorrect responses cause users to incur penalties that are both application-and item-specific, but current Bloom Filters are typically tuned only for static penalties. Such shortcomings are problematic for all Bloom Filter variants, but especially so for Time-decaying Bloom Filters, in which the memory of older items decays over time, causing both false positives and false negatives. We address these issues by introducing inferential filters, which integrate Bayesian priors and information latent in filters to make penalty-optimal, query-specific decisions. We also show how to properly infer insertion times in such filters. Our methods are general, but here we illustrate their application to inferential time-decaying filters to support novel query types and sliding window queries with dynamic error penalties. We present inferential versions of the Timing Bloom Filter and Generalized Bloom Filter. Our experiments on real and synthetic datasets show that our methods reduce penalties for incorrect responses to slidingwindow queries in these filters by up to 70% when penalties are dynamic.

Spatial databases are being used in an increasing number of application domains. Handling spatial... more Spatial databases are being used in an increasing number of application domains. Handling spatial joins efficiently has become a pressing problem, since most relational join methods are not directly applicable in the spatial domain. For example, the sort-merge join method requires a total order on join attributes values, which spatial attributes lack. Hash-join algorithms are efficient for natural joins, but spatial join criteria are too complex to be covered by the simple semantics of natural joins. Most current spatial join algorithms rely on pre-computed spatial indices, but this approach is inefficient and limited in its usefulness. Existing spatial indices are mostly designed for spatial selections, and are inefficient when used for joins. Further, the operand datasets of spatial joins do not always have pre-computed spatial indices, for example, when they are dynamically generated by other database operations. This dissertation proposes a new class of methods to support dynamic and efficient spatial join processing. Our work has two goals: first, to obviate the need for pre-computed spatial indices by designing spatial join methods that use dynamically constructed data structures, and second, to promote the efficiency of spatial join processing to the level of relational join processing. We propose a new spatial index, called the seeded tree, which is effective for spatial joins and efficient to construct at join time. Three principles guide the design of the seeded tree method. First, the construction of a join index should exploit available information about the join, so that the index is tailored for that join. We therefore guide the construction of seeded trees with information extracted either directly from the underlying dataset or an existing index tree. Second, the structure of join indices should be optimized for joins, and free of constraints imposed for supporting selection operations. By relaxing the structure of seeded trees to various degrees, we lower the construction cost and make join processing more efficient. Third, buffer space and I/O should be carefully managed. We develop the technique of batch writes that efficiently reclaims buffer space and accesses disks using sequential I/O. Our performance studies show that our methods run much faster than older methods. We then extend our spatial join methods, and propose the spatial hash-join approach, which combines the benefits of the relational hash-join paradigm and the seeded tree techniques. We examine the difficulties if applying relational techniques to spatial domains, and define a framework for designing spatial hash joins. Based on this framework, we design and test a spatial hash-join method, that requires no pre-computed indices, and performs better than current methods even when these methods are given pre-computed indices. This thesis provides very efficient solutions to the hitherto difficult and expensive problem of spatial joins. By not requiring pre-computed indices, it also increases the range of choices available to spatial query optimizers, enabling them to devise more efficient plans for complex queries. The buffer and I/O management techniques developed in this thesis can be applied to reduce I/O cost significantly for both spatial and relational join processing.

IEEE Computer Society Press eBooks, Mar 1, 1995

In this paper we describe the design and implementation of an integrated monitoring and debugging... more In this paper we describe the design and implementation of an integrated monitoring and debugging system for a distributed real-time computer system. The monitor provides continuous, transparent monitoring capabilities throughout a real-time system's lifecycle with bounded, minimal, predictable interference by using software support. The monitor is flexible enough to observe both high-level events that are operating system-and application-specific, as well as low-level events such as shared variable references. We present a novel approach to monitoring shared variable references that provides transparent monitoring with low overhead. The monitor is designed to support tasks such as debugging realtime applications, aiding real-time task scheduling, and measuring system performance. Since debugging distributed real-time applications is particularly difficult, we describe how the monitor can be used to debug distributed and parallel applications by deterministic execution replay. KEY WORDS Program monitoring Debugging Real-time systems 1 Distribution also imposes constraints on the monitor. Distributed systems lack both global state information and a sense of global time. There is no total ordering defined over events that occur on different nodes. Monitored data must be collected from several sites and integrated to obtain a coherent view of the system. Further,

International Journal of Sensor Networks, 2007

Broadcast data dissemination is well-suited for mobile wireless environments, where bandwidth is ... more Broadcast data dissemination is well-suited for mobile wireless environments, where bandwidth is scarce, and mutual interference must be minimized. However, broadcasting monopolizes the medium, precluding clients from performing any other communication. We address this problem in two ways. First, we segment the server broadcast, with intervening periods of silence, during which the wireless devices may communicate. Second, we reduce the average access delay for clients using a novel cooperative caching scheme. Our scheme is fully decentralized, and uses information available locally at the client. Our results show that our model prevents the server from monopolizing the medium, and that our caching strategy reduces client access delays significantly.

Nucleation and Atmospheric Aerosols, 1993

CIESIN (Consortium for International Earth Science Information Network) is funded by NASA to inve... more CIESIN (Consortium for International Earth Science Information Network) is funded by NASA to investigate the technology necessary to integrate and facilitate the interdisciplinary use of Global Change information. A clear part of this mission includes providing a link between the various global change data sets, in particular the physical sciences and the human (social) sciences. The typical scientist using the CIESIN system will want to know how phenomena in an outside field affects his/her work. For example, a medical researcher might ask: how does air-quality effect emphysema? This and many similar questions will require sophisticated semantic data integration. The researcher who raised the question may be familiar with medical data sets containing emphysema occurrences. But this same investigator may know little, if anything, about the existence or location of air-quality data. It is easy to envision a system which would allow that investigator to locate and perform a "join" on two data sets, one containing emphysema cases and the other containing air-quality levels. No such system exists today. One major obstacle to providing such a system will be overcoming the heterogeneity which falls into two broad categories. "Database system" heterogeneity involves differences in data models and packages. "Data semantic" heterogeneity involves differences in terminology between disciplines which translates into data semantic issues, and varying levels of data refinement, from raw to summary. Our work investigates a global data dictionary mechanism to facilitate a merged data service. Specifically, we propose using a semantic tree during schema definition to aid in locating and integrating heterogeneous databases.

International Journal of Machine Learning and Computing, 2011

This study proposes a genetic algorithm-based (GA-based) adaptive clustering protocol with an opt... more This study proposes a genetic algorithm-based (GA-based) adaptive clustering protocol with an optimal probability prediction to achieve good performance in terms of lifetime of network in wireless sensor networks. The proposed GA-based protocol is based on LEACH, called LEACH-GA herein, which basically has setup and steady-state phases for each round in the protocol and an additional preparation phase before the beginning of the first round. In the period of preparation phase, all nodes initially perform cluster head selection process and then send their messages with statuses of being a candidate cluster head or not, node IDs, and geographical positions to the base station. As the base station received the messages from all nodes, it then searches for an optimal probability of nodes being cluster heads via a genetic algorithm by minimizing the total energy consumption required for completing one round in the sensor field. Thereafter, the base station broadcasts an advertisement message with the optimal value of probability to the all nodes in order to form clusters in the following setup phase. The preparation phase is performed only once before the setup phase of the first round. The processes of following setup and steady-state phases in every round are the same as LEACH. Simulation results show that the proposed genetic-algorithm-based adaptive clustering protocol effectively produces optimal energy consumption for the wireless sensor networks, and resulting in an extension of lifetime for the network.

UMI, ProQuest ® Dissertations & Theses. The world's most comprehensive collectio... more UMI, ProQuest ® Dissertations & Theses. The world's most comprehensive collection of dissertations and theses. Learn more... ProQuest, Techniques for query optimization, scheduling, and load shedding in data stream management systems. ...

Google, Inc. (search). ...

Page 1. NView: A Visual Framework for Network Tool Integration * David G. Thaler and Chinya V. Ra... more Page 1. NView: A Visual Framework for Network Tool Integration * David G. Thaler and Chinya V. Ravishankar Electrical Engineering and Computer Science Department The University of Michigan, Ann Arbor, Michigan 48109-2122 t halerd@eecs.umich.edu Abstract ...

OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information), Sep 1, 1993