Sensor Planning for Bayesian Nonparametric Target Modeling

Bayesian nonparametric models, such as the Gaussian process and the Dirichlet process, have been extensively applied for target kinematics modeling in various applications including environmental monitoring, traffic planning, endangered species tracking, dynamic scene analysis, autonomous robot navigation, and human motion modeling. As shown by these successful applications, Bayesian nonparametric models are able to adjust their complexities adaptively from data as necessary, and are resistant to overfitting or underfitting. However, most existing works assume that the sensor measurements used to learn the Bayesian nonparametric target kinematics models are obtained a priori or that the target kinematics can be measured by the sensor at any given time throughout the task. Little work has been done for controlling the sensor with bounded field of view to obtain measurements of mobile targets that are most informative for reducing the uncertainty of the Bayesian nonparametric models. To present the systematic sensor planning approach to leaning Bayesian nonparametric models, the Gaussian process target kinematics model is introduced at first, which is capable of describing time-invariant spatial phenomena, such as ocean currents, temperature distributions and wind velocity fields. The Dirichlet process-Gaussian process target kinematics model is subsequently discussed for modeling mixture of mobile targets, such as pedestrian motion patterns.

Novel information theoretic functions are developed for these introduced Bayesian nonparametric target kinematics models to represent the expected utility of measurements as a function of sensor control inputs and random environmental variables. A Gaussian process expected Kullback Leibler divergence is developed as the expectation of the KL divergence between the current (prior) and posterior Gaussian process target kinematics models with respect to the future measurements. Then, this approach is extended to develop a new information value function that can be used to estimate target kinematics described by a Dirichlet process-Gaussian process mixture model. A theorem is proposed that shows the novel information theoretic functions are bounded. Based on this theorem, efficient estimators of the new information theoretic functions are designed, which are proved to be unbiased with the variance of the resultant approximation error decreasing linearly as the number of samples increases. Computational complexities for optimizing the novel information theoretic functions under sensor dynamics constraints are studied, and are proved to be NP-hard. A cumulative lower bound is then proposed to reduce the computational complexity to polynomial time.

Three sensor planning algorithms are developed according to the assumptions on the target kinematics and the sensor dynamics. For problems where the control space of the sensor is discrete, a greedy algorithm is proposed. The efficiency of the greedy algorithm is demonstrated by a numerical experiment with data of ocean currents obtained by moored buoys. A sweep line algorithm is developed for applications where the sensor control space is continuous and unconstrained. Synthetic simulations as well as physical experiments with ground robots and a surveillance camera are conducted to evaluate the performance of the sweep line algorithm. Moreover, a lexicographic algorithm is designed based on the cumulative lower bound of the novel information theoretic functions, for the scenario where the sensor dynamics are constrained. Numerical experiments with real data collected from indoor pedestrians by a commercial pan-tilt camera are performed to examine the lexicographic algorithm. Results from both the numerical simulations and the physical experiments show that the three sensor planning algorithms proposed in this dissertation based on the novel information theoretic functions are superior at learning the target kinematics with

little or no prior knowledge

Phosphorus export from a restored wetland ecosystem in response to natural and experimental hydrologic fluctuations

Wetland restoration is a commonly used approach to reduce nutrient loading to freshwater and coastal ecosystems, with many wetland restoration efforts occurring in former agricultural fields. Restored wetlands are expected to be effective at retaining or removing both nitrogen and phosphorus (P), yet restoring wetland hydrology to former agricultural fields can lead to the release of legacy fertilizer P. Here, we examined P cycling and export following rewetting of the Timberlake Restoration Project, a 440 ha restored riverine wetland complex in the coastal plain of North Carolina. We also compared P cycling within the restored wetland to two minimally disturbed nearby wetlands and an adjacent active agricultural field. In the restored wetland we observed increased soluble reactive phosphorus (SRP) concentrations following initial flooding, consistent with our expectations that P bound to iron would be released under reducing conditions. SRP concentrations in spring were 2.5 times higher leaving the restored wetland than a forested wetland and an agricultural field. During two large-scale drawdown and rewetting experiments we decreased the water depth by 1 m in ∼10 ha of inundated wetland for 2 weeks, followed by reflooding. Rewetting following experimental drainage had no effect on SRP concentrations in winter, but SRP concentrations did increase when the experiment was repeated during summer. Our best estimates suggest that this restored wetland could release legacy fertilizer P for up to a decade following hydrologic restoration. The time lag between restoration and biogeochemical recovery should be incorporated into management strategies of restored wetlands. Copyright 2010 by the American Geophysical Union.

A Mechanical Analysis of Suspensory Locomotion in Primates and Other Mammals

For primates, and other arboreal mammals, adopting suspensory locomotion represents one of the strategies an animal can use to prevent toppling off a thin support during arboreal movement and foraging. While numerous studies have reported the incidence of suspensory locomotion in a broad phylogenetic sample of mammals, little research has explored what mechanical transitions must occur in order for an animal to successfully adopt suspensory locomotion. Additionally, many primate species are capable of adopting a highly specialized form of suspensory locomotion referred to as arm-swinging, but few scenarios have been posited to explain how arm-swinging initially evolved. This study takes a comparative experimental approach to explore the mechanics of below branch quadrupedal locomotion in primates and other mammals to determine whether above and below branch quadrupedal locomotion represent neuromuscular mirrors of each other, and whether the patterns below branch quadrupedal locomotion are similar across taxa. Also, this study explores whether the nature of the flexible coupling between the forelimb and hindlimb observed in primates is a uniquely primate feature, and investigates the possibility that this mechanism could be responsible for the evolution of arm-swinging.

To address these research goals, kinetic, kinematic, and spatiotemporal gait variables were collected from five species of primate (Cebus capucinus, Daubentonia madagascariensis, Lemur catta, Propithecus coquereli, and Varecia variegata) walking quadrupedally above and below branches. Data from these primate species were compared to data collected from three species of non-primate mammals (Choloepus didactylus, Pteropus vampyrus, and Desmodus rotundus) and to three species of arm-swinging primate (Hylobates moloch, Ateles fusciceps, and Pygathrix nemaeus) to determine how varying forms of suspensory locomotion relate to each other and across taxa.

From the data collected in this study it is evident the specialized gait characteristics present during above branch quadrupedal locomotion in primates are not observed when walking below branches. Instead, gait mechanics closely replicate the characteristic walking patterns of non-primate mammals, with the exception that primates demonstrate an altered limb loading pattern during below branch quadrupedal locomotion, in which the forelimb becomes the primary propulsive and weight-bearing limb; a pattern similar to what is observed during arm-swinging. It is likely that below branch quadrupedal locomotion represents a “mechanical release” from the challenges of moving on top of thin arboreal supports. Additionally, it is possible, that arm-swinging could have evolved from an anatomically-generalized arboreal primate that began to forage and locomote below branches. During these suspensory bouts, weight would have been shifted away from the hindlimbs towards forelimbs, and as the frequency of these boats increased the reliance of the forelimb as the sole form of weight support would have also increased. This form of functional decoupling may have released the hindlimbs from their weight-bearing role during suspensory locomotion, and eventually arm-swinging would have replaced below branch quadrupedal locomotion as the primary mode of suspensory locomotion observed in some primate species. This study provides the first experimental evidence supporting the hypothetical link between below branch quadrupedal locomotion and arm-swinging in primates.