Electrical Engineering

Predicting the end-to-end packet delay and understanding the Internet dynamics are of great importance for many realtime and non-realtime applications, especially for provisioning the quality of service (QoS) of various sources. In this paper methods for modeling Internet dynamics and predicting the end-to-end delays are surveyed with emphasis on the approaches using system identification and time series analysis. The survey focuses on a control engineer's viewpoint rather than that of a statistician. Furthermore, it intends to provide a connection between the works in the both areas.

In this paper some parametric methods are introduced to characterize the self-similarity of DNA sequences. Compared with Fourier analysis, these methods perform statistically more stably and yield more reliable results. Using these methods, eight whole genomes of bacteria provided by NCBI are analyzed. Long-range correlation properties in the nucleotide density distribution along these DNA sequences are explored. Estimation results show that the long-range correlation structure prevails through the entire molecule of DNA. Higher order statistics through coarse graining reveal that rather than multifractal, there are only monofractal phenomena presented in the sequences. Hence, the nucleotide density distribution can be modeled asymptotically as fractional Gaussian noise. This result points to a new direction for analyzing and understanding the intrinsic structures of DNA sequences.

The problem of track-to-track association has been considered until recently in the literature only for pairwise associations. In view of the extensive recent interest in multisensor data fusion, the need to associate simultaneously multiple tracks has arisen. This is due primarily to bandwidth constraints in real systems, where it is not feasible to transmit detailed measurement information to a fusion center but, in many cases, only local tracks. As it has been known in the literature, tracks of the same target obtained from independent sensors are still dependent due to the common process noise [2]. This paper derives the exact likelihood function for the track-totrack association problem from multiple sources, which forms the basis for the cost function used in a multidimensional assignment algorithm that can solve such a large scale problem where many sensors track many targets. While a recent work [14] derived the likelihood function under the assumption that the track errors are independent, the present paper incorporates the (unavoidable) dependence of these errors.

Particle filtering has a great potential for solving highly nonlinear and non-Gaussian estimation problems, generally intractable within a standard linear Kalman filtering based framework. However, the implementation of particle filters (PFs) is rather computationally involved, which nowadays prevents them from practical real-world application. A natural idea to make PFs feasible for "real-time" data processing is to implement them on distributed multiprocessor computer systems. This paper presents three schemes for distributing the computations of generic particle filters, including resampling and, optionally, a Metropolis-Hastings (MH) step. Simulation results based on a maneuvering target tracking scenario show that distributed implementations can provide a promising solution to the steep computational burden incurred when using a large number of particles.

In this paper, we compare several nonlinear filtering methods, namely, extended Kalman filter (EKF), unscented filter (UF), particle filter (PF), and linear minimum mean square error (LMMSE) filter for a ballistic target tracking problem. We cast EKF and UF into a general linear recursive estimation framework and reveal their pros and cons. We pinpoint using the LMMSE filter for possible analytical solutions rather than starting with approximations such as system linearization or unscented transform. We compare the performance of EKF, UF, LMMSE filter and Gaussian PF for a ballistic target tracking problem. The estimation accuracy is also compared with the posterior Cramer-Rao lower bound (PCRLB). Our simulation results confirm that the LMMSE filter outperforms EKF and UF in terms of tracking accuracy, filter credibility and robustness against the sensitivity to filter initial condition. Its accuracy is slightly worse than that of Gaussian PF but with much lower computational load. We conclude that the LMMSE filter is preferred for the ballistic target tracking problem being studied.

In multi-sensor target tracking systems, measurements from the same target can arrive out of sequence, called the out-of-sequence measurements (OOSMs). The resulting problem -how to update the current state estimates with the "old" measurementshas been solved optimally and sub-optimally for onelag as well as multi-lag OOSM update. In general, the existing algorithms assume perfect target detection and no clutter in the received measurements. The real world has, however, possible missed target detection and random clutter in the possible OOSMs and thus the filter has to handle the measurement origin uncertainty. In this paper, we incorporate the probabilistic data association (PDA) into the two OOSM update algorithms ALG-I and ALG-II proposed previously. We present the algorithms ALG-I and ALG-II in new forms with economic storage and efficient computation based on the nonsingularity assumption of some special matrices. Simulation results show that PDA with the two OOSM update algorithms have compatible RMS errors to the in-sequence PDA filter.

The problem of track-to-track association and track fusion has been considered in the literature where the fusion center has access to multiple track estimates and the associated estimation error covariances from local sensors, as well as their crosscovariances. Due primarily to the communication constraints in real systems, some legacy trackers may only provide the local track estimates to the fusion center without any covariance information. In some cases, the local (sensor-level) trackers operate with fixed filter gain and do not have any self assessment of their estimation errors. In other cases, the network conveys a coarsely quantized root mean square (RMS) estimation error of each local tracker. Thus the fusion center needs to solve the track association and fusion problem with incomplete data from legacy local trackers.

Distributed Kalman filters are often used in multisensor target tracking where the fusion center receives local estimates and fuses them to obtain the global target state estimate. With such a fusion architecture, each local tracker can communicate less frequently with the fusion center than the local filter update rate. The global target state estimate via track fusion is usually less accurate than that of the centralized estimator when local estimation errors are correlated and local trackers communicate to the fusion center with bandwidth constraints lower than the measurement rate. This paper focuses on the tradeoff between bandwidth and tracking accuracy for track fusion with communication constraints. We show that the performance degradation increases for track fusion on demand compared with the centralized estimator as the number of local trackers increases. We relate the steady state analysis of track fusion under bandwidth constraints to noisy Wyner-Ziv source coding problem and compare our results with the theoretical rate distortion curve of the quadratic Gaussian CEO problem. We conclude that track fusion on demand is a side-information unaware strategy while the awareness of the correlated estimation errors at each local tracker can improve the track fusion accuracy significantly.