Keynote Presentations

Reliable target tracking requires the recursive estimation of dynamic object states from uncertain and heterogeneous sensor measurements. As the dimensionality of the state space, nonlinearities, and the number of possible measurement-to-track associations increase, conventional Bayesian filtering methods face substantial computational and memory challenges.

This talk explores the potential of quantum computing for target tracking and state estimation in multi-sensor data fusion. Following an introduction to Bayesian state estimation and probabilistic tracking, it discusses efficient density representations based on tensor decompositions and presents quantum concepts relevant to information fusion. Gate-based quantum algorithms are introduced for representing discretised state spaces and simulating drift and diffusion processes.

In addition, the presentation investigates adiabatic quantum computing for combinatorial tracking tasks such as data association, as well as energy-based formulations of Bayesian measurement updates. Quantum-inspired approaches, including wave-function and path-integral concepts, are also considered as novel methods for classical tracking applications.

Traditional identification methods for nonstationary linear systems fall into two categories: structured and unstructured methods. The first class, containing algorithms like LMS and RLS, does not assume any model of system parameters’ variation, which allows estimating parameters of relatively slowly varying systems. When system parameters vary rapidly, better estimation accuracy can be achieved by imposing a stochastic state-space model on coefficient evolution or by representing coefficient trajectories using a deterministic basis expansion, leading, for example, to Kalman filtering and basis-expansion approaches, respectively. However, the price of structured methods lies in increased computational complexity. An alternative is to use a hybrid approach, which applies unstructured methods to obtain approximately unbiased but raw estimates (called pre-estimates), which are further post-processed to obtain final estimates. This talk presents a family of pre-estimation techniques and discusses their properties and limitations. Finally, their application in self-interference cancellation in full-duplex underwater acoustic communication is presented.

This talk addresses recent developments in robust methods for analyzing cyclostationary signals. Cyclostationary models describe processes whose statistical characteristics vary periodically over time. The most widely studied class involves second-order cyclostationary processes, where second-order statistics exhibit periodicity. These signals are primarily characterized by their autocovariance (ACVF) or autocorrelation (ACF) functions. Traditionally, detecting cyclostationarity relies on evaluating the ACVF or ACF in either the time or frequency domain. In condition monitoring, the cyclic spectral coherence (CSC) map being a double Fourier transform of the ACVF, has become the standard tool for identifying cyclostationary features in the frequency domain. However, classical estimation techniques often break down when applied to real-world data. Real signals frequently exhibit non-Gaussian, heavy-tailed dynamics, resulting in prominent outliers that severely corrupt classical ACVF and ACF estimators and thus, also the ACVF and ACF-based techniques (like classical CSC map). To overcome these issues, recent cyclostationary signal processing techniques moves beyond Gaussian assumptions, introducing robust estimation frameworks designed for heavy-tailed environments. In this presentation, we discuss new strategies to preserve cyclostationary feature detection despite non-Gaussian noise, demonstrating their performance on vibration signals for local fault detection in mechanical systems.

This work is supported by the National Center of Science under the Sheng2 project No. UMO-2021/40/Q/ST8/00024 "NonGauMech - New methods of processing non-stationary signals (identification, segmentation, extraction, modeling) with non-Gaussian characteristics to monitor complex mechanical structures".

Neuromorphic engineering develops hardware and software systems inspired by biological neurons, with the goal of achieving energy-efficient, low-latency, robust, and adaptive computation, communication and control. Its potential impact on systems and control is significant, as it may enable novel approaches to control and estimation by leveraging brain-inspired computation and communication principles. In this context, we will see a framework for the robust stabilization of a plant subject to disturbances when the communication between noisy sensors and the controller relies on spiking signals generated by neuron inspired schemes. The communication scheme consists of a spike encoder on the sensors side, which is based on integrate-and-fire neurons that convert the analog plant output measurement into a spiking signal, and a spike decoder on the controller side inspired by synaptic processing to convert the received spiking signal into an analog signal. We will present design conditions on the spike decoder, the spike encoder as well as on the controller under which the closed-loop system exhibits a robust stability property, where the adjustable parameters are the amplitudes of the spikes. Numerical simulations on a single-link manipulator will be shown to illustrate the potential of the approach.