This paper presents a nonlinear filter to track a maneuvering target while accounting for a priori information about its maneuvering capabilities. The idea is to compute the probabilistic reachable set of the target given a priori distributions for its initial condition and maneuvering capabilities between two measurement time intervals. A non-product quadrature scheme, the Conjugate Unscented Transformation (CUT), is used to compute the statistical moments of the target state accurately with a minimal number of samples. A minimum variance framework is used to assimilate sensor observation with model prediction to estimate the target state and corresponding state covariance. The performance of the derived filter is benchmarked against a conventional sigma-point filter by considering the problem of tracking an automobile on a road network.

Feedback control problems pertaining to the control of tensegrity robotic systems are detailed in this paper. The unique problems that arise due to the positivity of the string tensions required to maintain the static stability and desirable stiffness of the structural system are shown to bring about interesting opportunities to optimize for the redundancy in the actuation process. The static stability consideration, coupled with the nonlinear dynamics and the sensor models introduce additional algebraic constraints in implementation of both kinematic and model based dynamic controllers for tensegrity systems. Approaches to develop kinematic and dynamic control techniques are detailed in this paper. A bench top experimental setup consisting of a simple tensegrity system is utilized to demonstrate the efficacy of the output feedback control approach developed in the paper. Near real time image measurements are utilized to drive the output error used in the control scheme.

A necessary step to building effective writing skill training system requires developing good methods for modeling human skill adequately. A number of groups have represented character information in various languages for writing skill based on trajectory information. These methods are often data intensive, tedious to implement and do not encode force information. To overcome these restrictions a novel present a novel modeling methodology that has good accuracy, robustness, flexibility and encodes force information involved in writing characters. This modeling methodology, based on Global-Local Approximation technique, has the capability to provide temporal force or position information independent of time, decoupling velocity information of the sample data used to capture skilled tasks. This modeling approach can be extended many human skilled tasks such as surgery, art and sports.

In contrast to traditional wireless sensor network (WSN) applications that perform only data collection and aggregation, the new generation of information processing applications such as pursuit-evasion games, tracking, evacuation, and disaster relief applications require in-network information storage and querying. Due to the resource limitations of WSNs, it is challenging to implement in-network querying in a distributed, lightweight, resilient, and energy-efficient manner. We address these challenges by exploiting location information and the geometry of the network and propose an in-network querying framework, namely, the Distributed Quad-Tree (DQT). DQT is distance sensitive for querying of an event: the cost of answering a query for an event is at most a constant factor (2radic(2) in our case) of the distance *d* to the event. DQT construction is local and does not require any communication. Moreover, due to its minimalist infrastructure and stateless nature, DQT shows graceful resilience to node failures and topology changes. Since event-based querying is inherently limited to the anticipated types of inquiries, we further extend our framework to achieve complex range-based querying. To this end, we use a multiresolution algorithm, which is optimal with respect to least square errors that models the data in a decentralized way. Our model-based scheme answers queries with approximate values accompanied by certainty levels with increased resolution at lower layers of the DQT hierarchy. Our analysis and experiments show that our framework achieves distance sensitivity and resiliency for event-based querying, as well as greatly reduces the cost of complex range querying.

Direction-dependent scaling, shaping, and rotation of Gaussian basis functions are introduced for maximal trend sensing with minimal parameter representations for input output approximation. It is shown that shaping and rotation of the radial basis functions helps in reducing the total number of function units required to approximate any given input-output data, while improving accuracy. Several alternate formulations that enforce minimal parameterization of the most general radial basis functions are presented. A novel "directed graph" based algorithm is introduced to facilitate intelligent direction based learning and adaptation of the parameters appearing in the radial basis function network. Further, a parameter estimation algorithm is incorporated to establish starting estimates for the model parameters using multiple windows of the input-output data. The efficacy of direction-dependent shaping and rotation in function approximation is evaluated by modifying the minimal resource allocating network and considering different test examples. The examples are drawn from recent literature to benchmark the new algorithm versus existing methods.