In this innovative practice work-in-progress paper, enrollment data from five institutions was used to examine equity in undergraduate research through Vertically Integrated Projects (VIP) Programs. VIP is a model for undergraduate research in which large student teams are embedded in faculty-driven projects. The American Association of Colleges and Universities recognizes undergraduate research as a high-impact experience, associated with higher graduation rates and greater learning gains in college. Participation in multiple high-impact experiences yields cumulative gains to students from all backgrounds, and compensatory gains for minoritized and marginalized students. Nationally however, minoritized students, first-generation college students, and transfer students participate in undergraduate research at lower rates than their peers. In this study, VIP enrollments at five institutions (N = 6,651 over two semesters) were compared to demographics of the institutions to determine the degree to which programs achieved equity among historically underserved minorities, transfer students, first-generation college students, and by gender. Analysis accounted for demographics and level of participation of the academic units involved, comparing enrollments with what would be expected under equitable enrollment. Analyses were done for each institution and across the pooled sample. By institution, equity across categories varied. Across the pooled sample, results show small effects sizes for status as a historically underserved minority, very small effect sizes for first-generation students and transfer students, and slightly higher participation among women than men. The large-scale nature of VIP teams enables institutions to scale-up their undergraduate research offerings. This paper begins answering the question of whether this scaling increases access for marginalized populations, and the results are encouraging. The paper is a work-in-progress, because data needs to be collected from more VIP institutions for a wider-ranging study. The chisquare test and the importance of using effect sizes in interpreting results will be explained, so others can apply the same method. Results, implications, and next steps are discussed.

The Level-up workshop will challenge exclusive and exclusionary models for undergraduate research experiences, and it will give participants tools to expand undergraduate research to serve all students. The model and associated tools are adaptable, and they have been implemented in 44 colleges and universities of varying sizes, settings and missions in 12 countries.

This innovative-practice work-in-progress paper explores student leadership development over multiple semesters in team-structured project-based courses. While student growth is expected in a single semester, the study asks if multiple semesters of participation lead to continued leadership growth, and if so, over how many semesters of participation growth continues. The study examined peer evaluation ratings in general leadership (coordination of teams’ work) and technical leadership (serving as a technical/content area leader) in a single semester of Georgia Tech’s Vertically Integrated Projects (VIP) Program, a multidisciplinary, multi-semester, team-structured, project-based, and credit-bearing program in which student teams support faculty research. Analysis examined means and distributions on two peer evaluation questions (N = 1,073 and N = 1,047) by student academic rank and number of semesters of participation in the program. Findings indicate that within their teams, students’ leadership increased through the third semester, with students making their greatest leadership contributions in the third semester and beyond; and students of lower academic rank provided as much leadership (including technical leadership) as older students who had comparable experience on the team. Both the VIP model and the operationalization of leadership represent innovative practices, because the VIP model yields measurable gains in student leadership, and the measurement of student leadership is based on peer-evaluations instead of self-assessments. The educational model and research in this paper are aligned with the FIE values of encouraging mentorship and professional growth, appreciating multidisciplinary approaches, valuing new approaches, and generating new knowledge. The paper addresses limitations and next steps for the study.

Wireless sensor and adhoc networks are hierarchically clustered for energy-efficiency, while gathering and aggregating data at the central clusterhead. Subsequent longrange communications from the clusterheads cause large-scale interference and energy-hole problems around them. It is thus better to have packets forwarded via short-range multi-hop routes between the clusterheads at different levels of the hierarchy. In order to discover the most optimal routes that serve the purpose, paths that minimize the inter-cluster routing delay within latticed clusters are analyzed. Consequently, a low-delay, energy-balancing distributed algorithm for routing across clusters is developed, which outperforms shortest path routing in high throughput networks. A parametric study comprising large-scale network evaluations is performed by developing an NS-3 based simulator.

Multi-level clustering offers the scalability that is essential to large-scale ad hoc and sensor networks in addition to supporting energy-efficient strategies for gathering data. The optimality of a multi-level network largely depends on two design variables: 1) The number of levels, and 2) The number of nodes operating at each level. We characterize these variables within a multi-hop, multi-level hierarchical network of variable size that gathers and aggregates data at each level. Our network communication cost model (EEHC-VA) is parameterized by the size of the data forwarded at each level, which depends on the application and aggregation strategy in place. We minimize the communication cost to obtain the optimal probabilities of distributed and independent selection of level-(n+1) nodes from level-n nodes. Interestingly, we have identified intervals-based on the number of nodes and aggregated data sizes-within which single- or two-level hierarchies are optimal. The results have been numerically verified for a wide range of parameters and validated with network simulations.

Localization is especially challenging in extreme RF emitter density (EED) environments (e.g. football stadiums), in part due to ambiguity in associating localization measurements to the correct emitter. One approach is to use other physical layer features for data association, but such techniques may not scale well for many emitters. This paper proposes exploiting the structure provided by the MAC layer for data association. The idea is explored in the context of IEEE 802.11g by using knowledge of the packet exchange sequence (PES), virtual carrier sense, and CSMA/CA to lower the probability of association error (PE) compared to a signal-to-noise ratio (SNR)-based OSI layer 1 strategy. Analytical expressions are derived for the PE on both a per packet detection and per packet exchange sequence basis. The proposed strategy lowers PE over an entire RTS/CTS sequence and scales well asymptotically in the number of emitters. The results are specific to WLANs, but the idea and approach are broadly applicable to any communications protocol with a MAC layer.

Multi-level clustering offers the scalability that is essential to large-scale ad hoc and sensor networks in addition to supporting energy-efficient strategies for gathering data. The optimality of a multi-level network largely depends on two design variables: 1) the number of levels and 2) the number of nodes operating at each level. We characterize these variables within a multi-hop multi-level hierarchical network of variable sizes that gathers and aggregates data at each level. Our network communication cost model (EEHC-VA) is parameterized by the size of the data forwarded at each level. We minimize the communication cost to obtain the optimal probabilities of distributed and independent selection of level-(n+1) nodes from level-n nodes. Interestingly, we have identified intervals-based on the number of nodes and aggregated data sizes-within which singleor two-level hierarchies are optimal. The results have been numerically verified for a wide range of parameters and validated with network simulations. Finally, the impact of these results on the network architectures is discussed for selected applications and aggregation schemes.

We consider a distributed detection problem in a large, single-hop, wireless sensor network. Because of limited collection time and bandwidth, the fusion center (FC) is not able to collect the local observations from all sensor nodes. A distributed detection scheme with a selection strategy and a capability to operate in a finite bandwidth is required. We propose an ordered sequential detection scheme which jointly integrates a reliability-based splitting algorithm, an ordered-transmission strategy, and a sequential probability ratio test (SPRT). The proposed scheme allows the FC to collect the local observations in descending order of their reliabilities by using a reliability-based splitting algorithm. As it receives successfully transmitted observations, the FC sequentially decides whether to make a global decision or to continue collecting more local observations. The numerical results show that the proposed scheme significantly outperforms a conventional SPRT scheme. The improvement increases as the number of sensor nodes in the network increases.

We determine the optimal nonuniform spatial density of nodes in a single-hop Wireless Sensor Network (WSN) whose task is the distributed detection of a target within its sensing field. The optimization approach accounts for such factors as the Medium Access Control (MAC) protocol being used, the wireless channel's propagation characteristics, a randomized sleep/wake-up scheduling protocol, network coverage constraints, the energy consumed, the time to reach a decision, and the number of nodes in the network. The node density that minimizes the average Decision Error Probability (DEP) when a node at the center of the network serves as the Cluster Head (CH) is shown to be a function of the distance from this CH. The solution of this optimization problem and simulations demonstrate both the significant performance improvement provided by nonuniform spatial densities and the trade-offs that are possible amongst energy, network lifetime, detection performance, and time to reach a decision.

We consider distributed detection applications for a fusion center that has a limited time to collect, weight and fuse local decisions to produce a global decision in a wireless sensor network that uses a random access protocol. When this time is not long enough to collect decisions from all nodes in the network, a strategy is needed for collecting those with the highest reliability. This is accomplished by incorporating a reliability-based splitting algorithm into the random access protocol: the collection time is divided into frames and only nodes with a specified range of reliabilities compete for the channel within each frame. Nodes with the most reliable decisions attempt transmission in the first frame; nodes with the next most reliable set of decisions attempt in the next frame; etc. To ensure that the results we derive are lower bounds on the detection performance that would be seen in practice, the local decisions that arrive within a frame are assigned the smallest reliability in the range associated with that frame. Two performance measures are used: Detection Error Probability (DEP) and Asymptotic Relative Efficiency (ARE) of the proposed scheme relative to a TDMA-based scheme. We then show how to minimize the DEP by determining the reliability intervals that define which nodes attempt to transmit in each frame. Intervals that maximize the channel throughput will often, but not always, minimize the DEP of the proposed scheme. Necessary conditions for the optimality of the maximum throughput intervals are thus derived. Numerical results indicate that these conditions are often sufficient when the observation noise is Gaussian. From the ARE, we determine when the proposed scheme asymptotically outperforms a TDMA-based scheme.

We consider a distributed detection application for a fusion center that has a limited time to make a global decision by collecting, weighting, and fusing local decisions made by nodes in a wireless sensor network that uses a random access channel. When this time is not long enough to collect decisions from all nodes in the network, a strategy is needed for collecting those with the highest reliability. This is accomplished with an easily implemented modification of the random access protocol: the collection time is divided into frames and only nodes with a particular range of reliabilities compete for the channel within each frame. Nodes with the most reliable decisions attempt transmission in the first frame; nodes with the next most reliable set of decisions attempt in the next frame; etc. We formulate an optimization problem that determines the reliability interval that defines who attempts in each frame in order to minimize the Detection Error Probability (DEP) at the fusion center. When the noise distribution affecting nodes' local decisions is continuous, symmetric, unimodal, and has a monotone likelihood ratio, reliability thresholds that maximize the channel throughput in each frame are optimal when the ratio of the collection time to the number of nodes is small. The number of frames that should be used depends on whether the average reliability or the worst-case reliability of local decisions in each frame is used to determine the DEP.

A sensor network's motes observe the environment, make estimates based on observations with spatially correlated noise sources, and then send/relay these estimates to a Cluster-Head (CH). A novel scheme based on dithered quantization and channel compensation is used to ensure that each mote's local estimate received by the CH is unbiased. Based on an upper bound of the noise covariance matrix, the CH fuses these unbiased local estimates into a global one using a Best Linear Unbiased Estimator (BLUE). We evaluate the mean square error(MSE) of the final estimate by both analysis and simulation.

In clustered networks of wireless sensors, each sensor collects noisy observations of the environment, quantizes these observations into a local estimate of finite length, and forwards them through one or more noisy wireless channels to the cluster head (CH). The measurement noise is assumed to be zero-mean and have finite variance, and each wireless hop is modeled as a binary symmetric channel (BSC) with a known crossover probability. A novel scheme is proposed that uses dithered quantization and channel compensation to ensure that each sensor's local estimate received by the CH is unbiased. The CH fuses these unbiased local estimates into a global one, using a best linear unbiased estimator (BLUE). Analytical and simulation results show that the proposed scheme can achieve much smaller mean square error (MSE) than two other common schemes, while using the same amount of energy. The sensitivity of the proposed scheme to errors in estimates of the crossover probability of the BSC channel is studied by both analysis and simulation. We then determine both the minimum energy required for the network to produce an estimate with a prescribed error variance and how this energy must be allocated amongst the sensors in the multihop network.

We use per-student virtual machines to allow new students to configure servers, thus enabling them to develop an understanding of the complex eStadium system. The outcomes include: student learning as the per-student virtual machines progress into software development and production machines supporting the eStadium game-day environment; the teamwork and leadership skills that evolve as students progress from initial learning to leadership roles in the creation of sophisticated applications; guidelines for instructors mentoring students through the process of building and maintaining a working production system; and, parallels with best-practice software and system development in industry. The use of peer-evaluations and social-network studies enable us to determine how the students interact with and learn from each other across years (sophomores through seniors). This cross year, cross experience-level learning process is essential for maintaining the technical and team continuity of the project. It also prepares students in a very realistic way for the software-development process in industry.

Georgia Tech's Colleges of Engineering and Computing initiated the Vertically-Integrated Projects (VIP) program in January 2009. Undergraduate students that join VIP teams earn academic credit for participating in design efforts that assist faculty and graduate students with research and development issues in their technical areas. The teams are: multidisciplinary - drawing students from around the university; vertically-integrated - maintaining a mix of sophomores through PhD students each semester; and long-term - each undergraduate student may participate in a project for up to six semesters. We describe the Video and Image Annotation VIP (VIA-VIP) project, which provides undergraduates unique opportunities to learn and apply state-of-the-art video-mining algorithms by processing a large archive of football videos recorded from GT football games. Their results are documented. Based on their feedback we believe the VIA-VIP course is on track to meet the needs of undergraduates in areas they don't usually see in the traditional undergraduate classroom.

The motes in a wireless sensor network observe the environment, make estimates based on their observations, and then send these estimates to a Cluster-Head (CH). A novel scheme based on dithered quantization and channel compensation is used to ensure that each mote's local estimate received by the CH is unbiased. Then the CH fuses these unbiased local estimates into a global one using a Best Linear Unbiased Estimator (BLUE). We determine the number of bits each node should transmit to reach a prescribed error variance at the CH subject to per-node limits on the energy that can be used and the per-node cost of forwarding a bit to the CH.

We study the problem of determining the optimal spatial node density for deployment of a Wireless Sensor Network (WSN) for distributed detection of a randomly located target in a sensing field. We formulate an optimization problem for the single cluster case and account for factors such as the Medium Access Control(MAC) protocol that is used, the wireless channel's propagation characteristics, a randomized sleep/wake-up scheduling protocol, network coverage constraints, the energy consumed, the time to reach a decision, and the total number of nodes available. We show that the optimal node density that minimizes the average Decision Error Probability (DEP) at the Cluster Head (CH) is a function of the distance from the CH and also examine trade-offs between the DEP and other network parameters as the spatial sensor density is varied across the cluster. Solution of the optimization problem and simulations are used to reveal the significant performance improvement that non-uniform spatial densities provide over uniform densities and the many optimal trade-offs that are possible amongst energy, network lifetime, detection performance and time to reach a decision.

We consider the problem of estimating the distance to a device transmitting in the 2.4 GHz ISM band that is interfering with users of an 802.11b wireless network. Accurate estimation of the distance enables a network designer to optimally configure transmit power levels and channel assignments of a wireless network. The estimate is made by a cluster of wireless sensor motes deployed along the edge of an 802.11b network. These motes perform a 1-bit quantization of the Received Signal Strength (RSS) using a dithered quantization framework. The quantized bits are transmitted over a Binary Symmetric Channel(BSC) to the Cluster Head (CH), which then uses a Maximum-Likelihood Estimation (MLE) technique to estimate the unknown parameter. We propose a framework in which the CH uses an iterative parameter estimation scheme in which it provides low overhead feedback to the motes to adjust their threshold values for the 1-bit dithered quantization process. Evaluation of the Root Mean Squared Error (RMSE) of this iterative scheme shows that it performs significantly better than iterative approaches in which all the motes use either identical thresholds or the non-identical thresholds proposed in. Our iterative scheme also more accurately tracks sudden changes in the distance to the interferer compared to previous approaches.

A sensor network's motes observe the environment, make estimates based on their observations, and send/relay these estimates to a Cluster-Head (CH). There are two sources of error in these multi-hop networks: observations are corrupted by noise and transmissions suffer communication errors. A novel scheme based on dithered quantization and channel compensation is used to ensure that each mote's local estimate received by the CH is unbiased. The CH fuses these unbiased local estimates into a global one using a Best Linear Unbiased Estimator (BLUE). We determine both the minimum energy required for the network to produce a BLUE estimate with a prescribed error variance and show how this energy should be allocated across the rings of a multi-hop network and the motes in each ring.

A set of mobile wireless sensors observe their environment as they move about. We consider the subset of these sensors that each made observations when they were all at approximately the same time/location. As they continue to move, one of them processes its observations and decides that an event that must be reported has taken place. To reduce the probability of a false alarm, this sensor assumes the role of a Cluster-Head (CH) and requests that all other sensors that collected observations at that time/location send it their decisions. The motion of each sensor determines how many hops its decision data must make to reach the CH. We analyze this effect of motion in the 1D case by modeling each sensor's motion as a Correlated Random Walk (CRW), which can account for transient behavior, geographical restrictions, and nonzero drift. Quantities, such as the energy required to collect the decision from all relevant sensors, can then be determined as a function of time.

To ensure that a multi-hop cluster of batterypowered, wireless sensor motes can complete all of its tasks, each task must minimize its use of communication and processing resources. For event detection tasks that are subject to both measurement errors by sensors and communication errors in the wireless channel, this implies that: (i) the Cluster-Head (CH) must optimally fuse the decisions received from its cluster in order to reduce the effect of measurement errors; (ii) the CH and all motes that relay other motes' decisions must adopt lowcomplexity processing and coding algorithms that minimize the effects of communication errors. This paper combines a Maximum a Posteriori (MAP) approach for local and global decisions in multi-hop sensor networks with low-complexity repetition codes and processing algorithms. It is shown by analysis and confirmed by simulation that there exists an odd integer M and an integer KMsuch the decision error probability at the CH is reduced when: (1) nodes in rings k ≤ KMhops from the CH directly relay their decisions to the CH; (2) nodes in rings k > KMlocally fuse groups of M decisions and then use a repetition code to forward these fused decisions to the CH; and (3) KMis a nondecreasing function of M. This algorithm - and hybrid, hierarchical, and compression approaches based on it - enable tradeoffs amongst the probability of error, energy usage, compression ratio, complexity, and time to decision.

In clustered networks of wireless sensor motes, each mote collects noisy observations of the environment, quantizes these observations into a local estimate of finite length, and forwards them through one or more noisy wireless channels to the Cluster Head (CH). The measurement noise is assumed to be zero-mean and have finite variance. Each wireless hop is assumed to be a Binary Symmetric Channel (BSC) with a known crossover probability. We propose a novel scheme that uses dithered quantization and channel compensation to ensure that each motes' local estimate received by the CH is unbiased. The CH then fuses these unbiased local estimates into a global one using a Best Linear Unbiased Estimator (BLUE). The energy allocation problem at each mote and among different sensor motes are also discussed. Simulation results show that the proposed scheme can achieve much smaller mean square error (MSE) than two other common schemes while using the same amount of energy. The sensitivity of the proposed scheme to errors in estimates of the crossover probability of the BSC channel is studied by both analysis and simulation.

We consider a multi-hop clustered wireless sensor network in which sensor nodes use their observations of the environment to make local hard (0/1) decisions about whether an event has occurred. The local decisions are transmitted over a binary symmetric channel (BSC) to the cluster head (CH) in a multi-hop fashion. The CH fuses the local decisions it receives to produce a final hard decision about the event. We investigate the asymptotic performance of the decision error probability (DEP) at the CH for a network with a fixed number of nodes and a randomly deployed network by deriving the error exponents. We analyze the performance improvement in the DEP and the error exponents when extra rings are added to the clustered network and show the effect of increasing the dimension of the network.

It has become clear that the original mobile Web model of people using their portable devices to surf normal Web- pages must be refocused on the creation of applications that are tailored for the mobile experience. The eStadium project is our attempt at creating such an application. It provides live "infotainment" such as real-time statistics, instant replay videos, and venue information to Purdue football fans via a public mobile Web application and an on- demand video delivery system. In this paper, we discuss the design, implementation, and operation of the eStadium system and the lessons learned from five years of serving real sports fans.

Communication among wireless sensor nodes that employ cheap low-power transceivers is often very sensitive to the variations of the wireless channel. Sensor network routing protocols thus strive to continually adapt to temporal variations in wireless links in order to avoid wasteful transmissions over low-quality links. Such adaptive routing protocols must rely on a scheme that can not only accurately estimate the quality of wireless links in terms of a quantitative measure, such as the packet success rate (PSR), but also quickly adapt to temporal dynamics of the links. Traditionally, the PSR is estimated from the fraction of successful transmissions over a window of test- packets. However, we demonstrate that counting based methods do not react to changes in the wireless channel fast enough and that the only way to address this problem is to estimate the PSR based on the receiver's characteristics and on the signal to noise ratio (SNR) at the receiver. We thus propose a scheme that uses a pre-calibrated SNR-PSR relationship and instantaneous SNR estimates to calculate the PSR of the link. In our scheme, each receiver continuously tracks the SNR using a Kalman Filter to minimize the estimation error and uses a locally available SNR- PSR curve to estimate the PSR. Through extensive experiments we demonstrate that our scheme adapts to variations in the channel faster than counting-based PSR estimators and that it also provides better PSR estimates than these counting-based approaches.

The stochastic model assumed to govern the mobility of nodes in a mobile ad hoc network has been shown to significantly affect the network's coverage, maximum throughput, and achievable throughput-delay trade-offs. In this paper, we compare several mobility models, including the random walk, random waypoint, and Manhattan models on the basis of the number of states visited in a fixed time, the time to visit every state in a region, and the effect of the number of wandering nodes on the time to first enter a set of states. These metrics for a mobility model are useful for assessing the achievable event detection rates in surveillance applications where wireless-sensor-equipped vehicles are used to detect events of interest in a city. We also consider mobility models based on Correlated Random Walks, which can account for time dependency, geographical restrictions, and nonzero drift. We demonstrate that these models are analytically tractable by using a matrix-analytic approach to derive new, closed-form results in both the time and transform-domains for the probability that a node is at any location at any time for both semi-infinite and finite 1D lattices. We also derive first entrance time distributions for these walks. We find that a correlated random walk 1) covers more ground in a given amount of time and takes a smaller amount of time to cover an area completely than a random walk with the same average transition rate, 2) has a smaller first entrance time to small sets of states than the random waypoint and random walk models, and 3) leads to a uniform distribution of nodes (except at the boundaries) in steady state.

In a clustered, multi-hop sensor network, a large number of inexpensive, geographically-distributed sensor nodes each make measurements of a source, quantize them into binary sequences, and transmit them over one or more wireless hops to the clusterhead. When all local measurement data has been gathered by the clusterhead, it fuses them into a final estimate about the source. Two sources of error affect the clusterhead's final estimate: (i) local measurement errors made by the sensor nodes because of noisy measurements or unreliable sensors; and (ii) bit errors affecting each hop on the wireless communication channel. Previous work assumed error-free communication or a single-hop cluster. We propose an optimal estimate that accounts for both of these sources of error. We show that this estimate significantly outperforms schemes that consider only the measurement error noise-both in terms of error counts and mean square error.

− In a clustered, multi-hop sensor network, a large number of inexpensive, geographically-distributed sensor nodes each use their observations of the environment to make local hard (0/1) decisions about whether an event has occurred. Each node then transmits its local decision over one or more wireless hops to the clusterhead. When all local decisions have been gathered by the clusterhead, it fuses them into a final hard decision about the event. Two sources of error affect the clusterhead’s final decision: (i) local decision errors made by the sensor nodes because of noisy measurements or unreliable sensors, and (ii) bit errors affecting each hop on the wireless communication channel. Previous work assumed error-free communication or a single-hop cluster. We show that if both of these sources of error are considered, then the optimal data fusion algorithm at the clusterhead is a weighted median. The optimal weights are shown to be functions of the bit error probability of the channel and the ring from which the local decision originated. We determine: the error probability of this optimal fusion algorithm; the effect of adding more nodes or rings to the cluster; and the tradeoff between energy consumed in the network and the decision error probability. This paper thus provides tools to add the effect of measurement and communication errors to other tradeoffs in the design of clustered sensor networks.

In a sensor cluster, a large number of geographically distributed sensor nodes each make local measurements and transmit them toward the clusterhead. The clusterhead may schedule simultaneous transmissions from nodes within the cluster by identifying non-interfering areas in the cluster. Directional antennas that are oriented toward the clusterhead may increase spatial reuse and thus increase the number of simultaneous transmissions. On the other hand, the antennas' narrower coverage areas may reduce connectivity, thus reducing the number of sensors from which data can be collected. In this paper, we investigate this tradeoff between spatial reuse and connectivity in the context of the spatiotemporal sampling rate that can be achieved by the cluster. We also propose a simple analytical approach to optimizing the antennas beam width in order to improve the spatiotemporal sampling rate by maximizing the transport capacity of the network

We consider scenarios in which an event, such as a mobile node arriving to collect data or a data-request arriving at a node via the sensor network, initiates the collection of one packet of data from each node in the surrounding cluster of nodes. The node requesting the data broadcasts its identity and purpose to the cluster and assumes the role of clusterhead. This broadcast message is also used to synchronize all nodes in the cluster, which covers an area with a radius of several hundred meters, to within a few microseconds. We exploit this ability to synchronize the cluster to improve retransmission algorithms in CSMA/CA-based MAC protocols. The result is a retransmission algorithm that uses a synchronized, shared contention window (SSCW) that is easy to implement and results in fewer collisions than retransmission algorithms that use binary-exponential-backoff (BEB). A simple analytical approach is further proposed to optimize the size of the shared contention window. We show via analytical, numerical, and simulation results that an SSCW-based 802.11 DCF protocol performs significantly better in these scenarios we consider than both 802.15.4 (ZigBee) and BEB-based 802.11 DCF (WiFi). We also show that SSCW-based 802.11 is more fault-tolerant and fair than TDMA

A multi-hop network of wireless sensors can be used to gather spatio-temporal samples of a physical phenomenon and transmit these samples to a processing center. This paper addresses an important issue in the design of such networks: determining the spatio-temporal sampling rate of the network under conditions of minimum energy usage. A new collision-free protocol for gathering sensor data is-used to obtain analytical results that characterize the tradeoffs among sensor density, energy usage, throughput, delay, temporal sampling rates and spatial sampling rates in wireless sensor networks. We also show that the lower bound on the delay incurred in gathering data is O(k(2)n) in a clustered network of n sensors with at most k hops between any sensor and its clusterhead (CH). Simulation results on the tradeoff between the achievable spatial sampling rates and the achievable temporal sampling rates when IEEE 802.11 distributed coordination function (DCF) is used as the media access scheme are provided and compared with the analytical results obtained in this paper.

A multihop network of wireless sensors can be used to gather spatio-temporal samples of a physical phenomenon and transmit these samples to a processing center. This paper addresses an important issue in the design of such networks: determining the spatio-temporal sampling rate of the network under conditions of minimum energy usage. Prior results on the design of energy-minimizing clustering algorithms and a new collision-free protocol for gathering sensor data are used to obtain analytical results that characterize the tradeoffs among sensor density, energy usage, temporal sampling rates and spatial sampling rates in wireless sensor networks. We also show that the lower bound on delay incurred in gathering data is O(k/sup 2/n) in a clustered network of n sensors with at most k hops between any sensor and its CH.

A wireless network consisting of a large number of small sensors with low-power transceivers can be an effective tool for gathering data in a variety of environments. The data collected by each sensor is communicated through the network to a single processing center that uses all reported data to determine characteristics of the environment or detect an event. The communication or message passing process must be designed to conserve the limited energy resources of the sensors. Clustering sensors into groups, so that sensors communicate information only to clusterheads and then the clusterheads communicate the aggregated information to the processing center, may save energy. In this paper, we propose a distributed, randomized clustering algorithm to organize the sensors in a wireless sensor network into clusters. We then extend this algorithm to generate a hierarchy of clusterheads and observe that the energy savings increase with the number of levels in the hierarchy. Results in stochastic geometry are used to derive solutions for the values of parameters of our algorithm that minimize the total energy spent in the network when all sensors report data through the clusterheads to the processing center.

This paper develops a technique for quantifying and minimizing the energy required to gather data from all sensors in a multi-hop wireless sensor network The technique combines analytical results from stochastic geometry with a distributed, randomized algorithm for generating clusters. The minimum communication energy achieved by this approach is significantly lower than the energy costs incurred in non-clustered networks and in clustered networks produced by such algorithms as the max-min d-cluster algorithm.

Stack filters are a class of nonlinear filters with excellent properties for signal restoration. Unfortunately, present algorithms for designing stack filters can only be used for small window sizes because of either their computational overhead or their serial nature. This paper presents a new adaptive algorithm for determining a stack filter that minimizes the mean absolute error criterion. The new algorithm retains the iterative nature of many current adaptive stack filtering algorithms, but significantly reduces the number of iterations required to converge to an optimal filter. This algorithm is faster than all currently available stack filter design algorithms, is simple to implement, and is shown in this paper to always converge to an optimal stack filter. Extensive comparisons between this new algorithm and all existing algorithms are provided. The comparisons are based both on the performance of the resulting filters and upon the time and space complexity of the algorithms. They demonstrate that the new algorithm has three advantages: it is faster than all other available algorithms; it can be used on standard workstations (SPARC 5 with 48 MB) to design filters with windows containing 20 or more points; and, its highly parallel structure allows very fast implementations on parallel machines. This new algorithm allows cascades of stack filters to be designed; stack filters with windows containing 72 points have been designed in a matter of minutes under this new approach.

We introduce a system for machine recognition of music patterns. The problem is put into a pattern recognition framework in the sense that an error between a target pattern and scanned pattern is minimized. The error takes into account pitch and rhythm information. The pitch error measure consists of an absolute (objective) error and a perceptual error. The latter depends on an algorithm for establishing the tonal context which is based on Krumhansl's (1990) key-finding algorithm. The sequence of maximum correlations that it outputs is smoothed with a cubic spline and is used to determine weights for perceptual and absolute pitch errors. Maximum correlations are used to create the assigned key sequence, which is then filtered by a recursive median filter to improve the structure of the output of the key finding algorithm. A procedure for choosing weights given to pitch and rhythm errors is discussed.

The present approach to the MAE-based design of stack filters for image restoration does not always produce the desired visual result. Thus, in this paper, a new stack filter design algorithm is developed. It is based upon a weighted mean absolute error (WMAE) criterion instead of the traditional MAE criterion, which assigns the same weights to all errors. The weights in this WMAE criterion are designed with the aid of the visible differences predictor (VDP), which can estimate the sensitivity of the human visual system to changes in images. Experiments with this WMAE approach show that the stack filters it produces perform significantly better in image processing applications than those designed with the MAE approach.

One of the most challenging problems in engineering-and signal processing-education is providing realistic and meaningful design experience. In the Engineering Projects in Community Service (EPICS) program, teams of engineering undergraduates earn academic credit for multi-year projects that solve technology-based problems for community organizations. Key features of EPICS include the long-term nature of the projects; emphasis on "real-world" start-to-finish design; the learning experience embodied in solving ambitious engineering problems; vertical, multidisciplinary teams; development of teamwork and communication skills; and the use of engineering to help the community. We describe the EPICS program and highlight four EPICS signal processing projects: a real-time system to measure speaking rate for Purdue's speech clinic; voice-controlled interactive software to encourage speech in developmentally delayed children; a microphone array hearing aid; and a virtual museum tour and interactive Web-based history games for the Tippecanoe County Historical Association.

The present approach to the MAE-based design of stack filters for image restoration does not always produce the desired visual result. Thus, in this paper, a new stack filter design algorithm is developed. It is based upon a weighted mean absolute error (WMAE) criterion instead of the traditional MAE criterion that assigns the same weights to all errors. The weights in the WMAE criterion are designed with the visible differences predictor (VDP), which can estimate the sensitivity of the human visual system to changes in images. Experiments with this WMAE approach show that the stack filters it produces perform significantly better in image processing applications than those designed with the MAE approach.

The theory of optimal stack filtering has been used in the difference of estimates (DoE) approach to the detection of intensity edges in noisy images. The DoE approach is modified by imposing a symmetry condition on the data used to train the two stack filters. Under this condition, the stack filters obtained are duals of each other. Only one filter must therefore be trained; the other is simply its dual. This new technique is called the symmetric difference of estimates (SDoE) approach. The dual stack filters obtained under the SDoE approach are shown to be comparable. This allows the difference of these two filters to be represented by a single equivalent edge operator. This latter result suggests that an edge operator can be found by directly training a (possibly nonpositive) Boolean function to be used on each level of the threshold decomposition architecture. This approach, which is called the threshold Boolean filter (TBF) approach, requires less training time but produces operators that are less robust than those produced by the SDoE approach. This is demonstrated and interpreted via comparisons of results for natural images.

We are developing a new graduate program at Purdue in Video and Image Systems Engineering (VISE). The project is comprised of three parts: a new curriculum centered around a degree option in VISE to be earned as part of the Masters or Ph.D. degrees; a state-of-the-art lecture/laboratory facility for instruction, laboratory experiments, and project and homework activities in VISE courses; and enhancement of existing courses and development of new courses in the VISE area.

In this paper we describe two error-recovery approaches for MPEG encoded video over ATM networks. The first approach aims at reconstructing each lost pixel by spatial interpolation from the nearest undamaged pixels. The second approach recovers lost macroblocks by minimizing intersample variations within each block and across its boundaries. Moreover, a new technique for packing ATM cells with compressed data is also proposed.

The transient behavior of time-dependent M/M/1 queues is studied. The boundary probability function pi /sub 0/(t), which is the probability that the queue is empty at time t, is shown with analyticity arguments to satisfy a Volterra-type integral equation. The boundary integral equation is derived, and a numerical algorithm is used to solve the integral equation and to find the expected queue size from pi /sub 0/(t). The approach can be applied to many other types of time-dependent queues. Examples are given.>

Quasi-birth-death (QBD) processes are used as performance models for voice/data integration schemes. It is assumed that voice and data packets of different bandwidths require different numbers of servers for service. Voice packets are lost when there are no available servers; data packets are queued in a buffer when there are not enough servers. A new approach is proposed to find the transient solutions of the QBD processes. The method is applied to a QBD model of a voice/data integration scheme to find time-dependent performance measures such as the expected number of data or voice packets in the system and the voice packet blocking probability.>

A new optimization theory for stack filters is presented in this paper. This new theory is based on the minimax error criterion rather than the mean absolute error (MAE) criterion used in [8]. In the binary case, a methodology will be designed to find the stack filter that minimizes the maximum absolute error between the input and the output signals. The most interesting feature of this optimization procedure is the fact that it can be solved using a linear program (LP), just like in the MAE case [8]. One drawback of this procedure is the problem of randomization due to the lost of structure in the constraint matrix of the LP. Several sub-optimal solutions will be discussed and an algorithm to find an optimal integer solution (still using a LP) under certain conditions will be provided. When generalizing to multiple-level inputs, complexity problems will arise and two alternatives will be suggested. One of these approaches assumes a parameterized stochastic model for the noise process and the LP is to pick the stack filter which minimizes the worst effect of the noise on the input signal.

Progress is reported in the design and analysis of spread-spectrum packet radio networks for the factory of the future. This progress includes the accurate analysis of the probability of packet success in a direct-sequence spread-spectrum communication channel and analytical results on the transient behavior of packet radio networks. The first use of these networks will be to provide communications between a transport system controller and a fleet of autonomous guided vehicles.>

A class of sliding window operators called generalized stack filters is developed. This class of filters, which includes all rank order filters, stack filters, and digital morphological filters, is the set of all filters possessing the threshold decomposition architecture and a consistency property, called the stacking property. A linear program is provided which determines a generalized stack filter which minimizes the Mean Absolute Error (MAE) between the output of the filter and a desired input signal, given noisy observations of that signal. These results show that choosing the generalized stack filter which minimizes the MAE is equivalent to massively parallel threshold-crossing decision making when these decisions are consistent with each other.>

A continuous-time Markov chain model of the CSMA/CD (carrier-sense multiple-access with collision detection) protocol is used to find its delay characteristics. For both random-service-order (RSO) and first-come-first-serve (FCFS) CSMA/CD networks, matrix-geometric methods are used to derive and/or compute the throughput-delay and caudal-characteristics curves, the distribution and variance of the delay, and the rate of decay of the tail of the delay distribution. These results can be used to determine the suitability of RSO and FCFS CSMA/CD networks in applications in which they must carry delay-sensitive data. This is illustrated by an example which shows the significant improvement obtained in the delay characteristics of a particular network when FCFS is used instead of RSO.>

With the aid of the threshold decomposition, rank order filters are shown to be the optimal filters under a general error criterion which includes an absolute error criterion. The particular rank order filter that is optimal is specified by a linear program. The technique developed in this paper extends to an entire class of nonlinear filters known as stack filters [5].

The properties of the median filter can be used to great advantage in the improvement of television images corrupted with "sparklie" noise. In this paper we describe the application of the median filter to a five second noisy movie sequence received from a satellite antenna with a weak signal. The sequence of two dimensional TV images is decomposed into a set of one dimensional signals. Each one dimensional signal consists of the succession of values a fixed picture element (pixel) takes as successive frames arrive for display. The median filter is applied separately to each of these one dimensional signals. It is shown that the portion of TV frames not containing noise almost always appears as a root signal to the median filter. Finally a real time implementation of the filtering scheme is proposed.

A continuous time Markov chain model of asynchronous, nonpersistent CSMA/CD networks is developed which allows their complete characterization by a matrix whose dimensions are independent of the number of users in the network. The model provides an efficient reeursive algorithm for computing the performance of finite population networks. It also allows many analytical results to be obtained when an infinite population and a constant load retransmission policy are considered. These analytic results include necessary and sufficient conditions for stability, the distribution of the number of busy users, an exact closed form expression for the expected delay, and the exact upper bound on the throughput-all obtained via the properties of the matrix mentioned above. The results on stability show that, for this CSMA/CD model, the definitions of stability in terms of expected drift, nonnull recurrence, and maximum possible throughput exceeding actual loading are all equivalent. The model can approximate as closely as desired any distribution for the packet length, transmission detection time, and collision truncation time. Taking advantage of this freedom, the performance of CSMA/CD networks is shown to be very sensitive to the distribution of the transmission detection time.

This paper describes the derivation and verification of a phoneme model of English speech. The model is used to generate a stream of phonemically labeled speech frames to model speech input for the design of a distributed speech understanding system. New computer architectures to perform speech understanding in real time should incorporate information about the characteristics of English speech. In order to predict the performance of a new architecture, it is necessary to simulate the design using either massive amounts of speech data or, as an alternative, a statistical model of speech. A statistically generated phoneme stream is used to avoid the difficulty of performing computationally intensive acoustic parameterization on the enormous amount of speech input data which would be required to obtain representative phoneme distributions and patterns of speech.

A continuous time Markov chain model of a nonpersistent carrier sense multiple access with collision detection (CSMA/CD) protocol is developed to determine the performance of finite population systems. The model allows a wide range of distributions for the length of packets and the lengths of the transmission detection and collision detection times. A stability measure for finite population asynchronous systems-the instantaneous expected drift-is developed. Performance calculations for many sample systems show that the following behavior is characteristic of these systems: a wide range of retransmission rates provides almost optimal performance; the instantaneous drift is a sensitive indicator of the stability characteristics of CSMA/CD networks; if high fixed retransmission rates are used, performance decays very rapidly when other network parameters cross critical thresholds. In addition, adaptive retransmission policies and the maximum length of a CSMA/CD cable network are found to be very sensitive to the distribution of the transmission detection time.