To extend existing distance metrics in the $L^{p}$ space, we define a novel distance, named relative Euclidean distance (RED), and prove that it is positive definite, symmetric, and satisfies the triangle inequality. This distance has a value range of [0, 1]. We prove a uniqueness of it in the $L^{p}$ space. The technique for order preference by similarity to ideal solution (TOPSIS) is widely used for multiple-attribute decision problems. The commonly used relative closeness measure in the TOPSIS is not a mathematical distance and thus does not have all the nice properties of such a distance. To improve the TOPSIS, particularly its relative closeness measure, we propose two measures, named optimistic distance (OD) and pessimistic distance (PD), based on our RED, and use them to measure the relative closeness in the TOPSIS. Both measures are positive definite, symmetric, and satisfy the triangle inequality. The best choice in OD (or PD) is on the Pareto frontier. Three examples of performance ranking are given to illustrate the usefulness of the TOPSIS with our proposed measures.
31 August 2017 in IEEE Transactions on Aerospace and Electronic Systems
This paper deals with target classification by using both feature data and kinematic measurements. The problem is tackled by multi-hypothesis sequential testing with embedded target tracking.We implement an Armitage sequential test for non-maneuvering and maneuvering targets. Both (centralized and distributed) fusion architectures are used for the embedded tracking. The contributions of the kinematic measurements to classification are analyzed, and classification performance improvement is shown analytically for a special case. Numerical results are provided to demonstrate the performance of our algorithms.
This paper formulates and studies the problem of distributed filtering based on randomized gossip strategy in order to estimate the state of a dynamic system via all sensors in a network. First we introduce the randomized gossip algorithm by which the fastest averaging strategy can be obtained for a network with an arbitrary topology. Then we combine the randomized gossip algorithm with the information filter to design a randomized gossip based distributed filtering algorithm. The proposed method can adopt different communication volume flexibly, which results in different estimation performance. This flexibility distinguishes our method from the existing ones. Simulation examples verify that our method outperforms the diffusion strategy based distributed filtering algorithm if a small increase of communication requirements is allowed.
This paper studies distributed state estimation for a dynamic system based on measurements collected by every node in a sensor network. Depending on the communications among nodes, information spreads across the network node by node through iterations. Then, nodes work collaboratively to estimate the system state timely. Most existing networked filters aim to make all nodes reach consensus at the centralized estimation with sufficient communications. To accommodate limited communication resources in practice, this paper proposes to optimize the entire network's estimation accuracy given a fixed number of communications. This is done by optimizing the sequence of node activations for selective communication. We propose two selective activation schemes: link activation (LA) and star activation (SA). In each iteration, they activate a single link and a single star (i.e., a node with all its neighbors), respectively. We develop two iterative distributed filters (DF): LA based (LA-DF) and SA based (SA-DF). LA-DF and SA-DF possess many important properties. For example, they are unbiased, convergent, stable, and credible. Finally, we analyze the performance of our filters and provide simulation results compared with existing filters to verify the superiority of the proposed filters.
Belief propagation (BP) is widely used to solve the cooperative localization problem due to its excellent performance and natural distributed structure of implementation. For a mobile agent network, its factor graph inevitably encounters loops. In this case, the BP algorithm becomes iterative and can only provide an approximate marginal probability density function of the estimate with finite iterations. We propose an augmented-state BP algorithm for mobile agent networks to alleviate the effect of loops. By performing state augmentation, the messages in the factor graph will actually be allowed to be backward propagated, which reduces the number of loops in the factor graph, increases the available information of agents, and thus, benefits the localization. Experimental results demonstrate the better performance of the proposed algorithm over the original BP method.