The aim of developing bio-inspired sensing systems is to try and emulate the amazing sensitivity and specificity observed in the natural world. These capabilities have evolved, often for specific tasks, which provide the organism with an advantage in its fight to survive and prosper. Capabilities cover a wide range of sensing functions including vision, temperature, hearing, touch, taste and smell. For some functions, the capabilities of natural systems are still greater than that achieved by traditional engineering solutions; a good example being a dog's sense of smell. Furthermore, attempting to emulate aspects of biological optics, processing and guidance may lead to more simple and effective devices. A bio-inspired sensing system is much more than the sensory mechanism. A system will need to collect samples, especially if pathogens or chemicals are of interest. Other functions could include the provision of power, surfaces and receptors, structure, locomotion and control. In fact it is possible to conceive of a complete bio-inspired system concept which is likely to be radically different from more conventional approaches. This concept will be described and individual component technologies considered.
It is common practice to represent a target group (or an extended target) as set of point sources and attempt to formulate a tracking filter by constructing possible assignments between measurements and the sources. We suggest an alternative approach that produces a measurement model (likelihood) in terms of the spatial density of measurements over the sensor observation region. In particular, the measurements are modelled as a Poisson process with a spatially dependent rate parameter. This representation allows us to model extended targets as an intensity distribution rather than a set of points and, for a target formation, it gives the option of modelling part of the group as a spatial distribution of target density. Furthermore, as a direct consequence of the Poisson model, the measurement likelihood may be evaluated without constructing explicit association hypotheses. This considerably simplifies the filter and gives a substantial computational saving in a particle filter implementation. The Poisson target-measurement model will be described and its relationship to other filters will be discussed. Illustrative simulation examples will be presented.
We propose a very generic Bayesian framework for the principled exploitation of probabilistic batch-learning technologies for real-time state estimation. To illustrate our concepts, we derive a nonlinear filtering/smoothing solution for a challenging case study in target tracking. We also demonstrate the application of Markov chain Monte Carlo (MCMC) sampling methods as a computational tool within our framework. Finally, we present simulation results, benchmarked against a comparable particle filter.
The problem of maintaining track on a primary target in the presence spurious objects is addressed. Recursive and batch filtering approaches are developed. For the recursive approach, a Bayesian track splitting filter is derived which spawns candidate tracks if there is a possibility of measurement misassociation. The filter evaluates the probability of each candidate track being associated with the primary target. The batch filter is a Markov-chain Monte Carlo (MCMC) algorithm which fits the observed data sequence to models of target dynamics and measurement-track association. Simulation results are presented.
For many dynamic estimation problems involving nonlinear and/or non-Gaussian models, particle filtering offers improved performance at the expense of computational effort. This paper describes a scheme for efficiently tracking multiple targets using particle filters. The tracking of the individual targets is made efficient through the use of Rao-Blackwellisation. The tracking of multiple targets is made practicable using Quasi-Monte Carlo integration. The efficiency of the approach is illustrated on synthetic data.
We consider the problem of tracking a group of point targets via a sensor with limited resolution and a finite field of view. Measurement association uncertainty and measurement process non-linearity are major difficulties with such cases. It is shown that a Bayesian estimator can be directly implemented using the particle filter technique.
The problem of tracking point targets moving in a group, or features of an extended object, is formulated via a general two component model. An example involving translation, scaling, rotation and pattern distortion is presented. It is assumed that measurements of the points are unlabelled, which, together with a significant clutter level, leads to measurement association uncertainty. A Bayesian bootstrap filter is used to implement a nonlinear, multiple hypothesis, recursive estimator.
The sampling based bootstrap filter is applied to the problem of maintaining track on a target in the presence of intermittent spurious objects. This problem is formulated in a multiple hypothesis framework and the bootstrap filter is applied to generate the posterior distribution of the state vector of the required target - i.e. to generate the target track. The bootstrap technique facilitates the integration of the available information in a near-optimal fashion without the need to explicitly store and manage hypotheses from previous time steps.
A Bayesian technique is applied to the target acquisition problem at handover from the fire control to the missile seeker. This is a multiple hypothesis scheme which includes an explicit model of the possible misalignment or bias between the fire control and seeker coordinate frames. This method is compared with a linear least cost assignment technique which may be implemented via the Munkres fast search algorithm.
A standard assumption of most multiple target tracking filters is that all the targets move independently of one another. However, in many cases, it is known a-priori that the targets move (at least approximately) as a group: this dependency should not be ignored. In this paper we describe an approach to multiple target tracking where the target dynamics are taken to be the superposition of a group effect which is common to all group members and an individual effect which is taken to be independent between members of the group. The method also allows for the presence of clutter and missed target detections. This is done by embedding the dependent target motion model within a multiple hypothesis framework. A closed form solution is derived for the special linear-Gaussian case and simulation results illustrating performance are presented. This paper is a recursive extension of the selection method presented at last year's conference.
The following problem is considered: a group of point targets is observed via an imperfect sensor and one of the measurements chosen. The measurements of each target position is corrupted by an independent error, although every object is detected. Two processes then act to move and distort the group: one is a bulk effect that acts equally on all members of the group while the other is independent for each target. The group is observed again by a (possibly different) imperfect sensor which may not detect every target. The problem is to construct the posterior distribution of the chosen target's position, given the two sets of measurements. Probability models of the sensors and of the pattern distortion processes are assumed to be available. A formal general solution has been obtained for this problem. For the special linear-Gaussian case this reduces to a closed form analytic expression. To facilitate implementation, a hypothesis pruning technique is given. A simulation example illustrating performance is provided.
The Bayesian solution of the problem of tracking a target in random clutter gives rise to Gaussian mixture distributions, which are composed of an ever increasing number of components. To implement such a tracking filter, the growth of components must be controlled by approximating the mixture distribution. A popular and economical scheme is the Probabilistic Data Association Filter (PDAF), which reduces the mixture to a single Gaussian component at each time step. However this approximation may destroy valuable information, especially if several significant, well spaced components are present. <p> </p>In this paper, two new algorithms for reducing Gaussian mixture distributions are presented. These techniques preserve the mean and covariance of the mixture, and the fmal approximation is itself a Gaussian mixture. The reduction is achieved by successively merging pairs of components or groups of components until their number is reduced to some specified limit. Further reduction will then proceed while the approximation to the main features of the original distribution is still good. <p> </p>The performance of the most economical of these algorithms has been compared with that of the PDAF for the problem of tracking a single target which moves in a plane according to a second order model. A linear sensor which measures target position is corrupted by uniformly distributed clutter. Given a detection probability of unity and perfect knowledge of initial target position and velocity, this problem depends on only twç non-dimensional parameters. Monte Carlo simulation has been employed to identify the region of this parameter space where significant performance improvement is obtained over the PDAF.