This technical note introduces the closed form maximum likelihood estimator for estimating the coefficients of the non-parametric frequency response function from system identification experiments. It is assumed that the experiments consist of repeated pulse excitations and that both the excitation and system response are measured which leads to an error-in-variables setting. Monte Carlo simulations indicate that the estimator achieves efficiency at low signal-to-noise ratios with only few measurements. Comparison with the least-squares estimator shows that better, unbiased results are obtained.

In this article, a two filter particle smoothing algorithm for Wiener state-space systems is proposed. The smoother is obtained by exploiting the model structure. This leads to a suitable proposal density for the backward filter inherent in the problem instead of introducing an artificial one. Numerical examples are provided in order to illustrate the proposed algorithm's performance and to compare it to current state of the art smoothers from the literature. It is found that the proposed method yields comparable results with less computational complexity as backward simulation-based particle smoothing algorithms.

The main contribution of this paper is a comparison of different machine learning algorithms for vehicle classification according to the "Nordic system for intelligent classification of vehicles" standard using measurements of road surface vibrations and magnetic field disturbances caused by vehicles. The algorithms considered are logistic regression, neural networks, and support vector machines. They are evaluated on a large dataset, consisting of 3074 samples and hence, a good estimate of the actual classification rate is obtained. The results show that for the considered classification problem logistic regression is the best choice with an overall classification rate of 93.4%.

This article shows how a particle smoother based system identification method can be applied for estimating the trajectory of road vehicles. As sensors, a combination of an accelerometer measuring the road surface vibrations and a magnetometer measuring magnetic disturbances mounted on the side of the road are considered. First, sensor models describing the measurements of the two sensors are introduced. It is shown that these depend on unknown, static parameters that have to be considered in the estimation. Second, the sensor models are combined with a two-dimensional constant velocity motion model. Third, the system identification algorithm is introduced which iteratively runs a Rao-Blackwellized particle smoother to estimate the vehicle trajectory followed by an expectation-maximization step to estimate the parameters. Finally, the method is applied to both simulation and measurement data. It is found that the method works well in general and some issues when real data is considered are identified as future work.

In this paper, we propose a method for vehicle tracking on roadways using measurements of magnetometers and accelerometers. The measurements are used to build a low-cost, low-complexity vehicle tracking sensor platform for highway traffic monitoring. First, the problem is formulated by introducing the process model for the motion of the vehicle on the road and two measurement models: one for each of the sensors. Second, it is shown how the measurements of the sensors can be fused using particle filtering. The standard sampling importance resampling (SIR) particle filter is extended for processing of multirate sensor measurements and models that employ unknown static parameters. The latter are treated by Rao–Blackwellization. The performance of the method is demonstrated by computer simulations. It is found that it is feasible to fuse the two sensors for vehicle tracking and that the proposed multirate particle filter performs better than particle filters that process only measurements of one of the sensors. The main contribution of this paper is the novel approach of fusing the measurements of road-mounted magnetometers and accelerometers for vehicle tracking and traffic monitoring.

Accurate and robust mobile robot localization is very important in many robot applications. Monte Carlo localization (MCL) is one of the robust probabilistic solutions to robot localization problems. The sensor model used in MCL directly influence the accuracy and robustness of the pose estimation process. The classical beam models assumes independent noise in each individual measurement beam at the same scan. In practice, the noise in adjacent beams maybe largely correlated. This will result in peaks in the likelihood measurement function. These peaks leads to incorrect particles distribution in the MCL. In this research, an adaptive sub-sampling of the measurements is proposed to reduce the peaks in the likelihood function. The sampling is based on the complete scan analysis. The specified measurement is accepted or not based on the relative distance to other points in the 2D point cloud. The proposed technique has been implemented in ROS and stage simulator. The result shows that selecting suitable value of distance between accepted scans can improve the localization error and reduce the required computations effectively.

Traffic monitoring using low-cost two-axis magnetometers is considered. Although detection of metallic vehicles is rather easy, detecting the driving direction is more challenging. We propose a simple algorithm based on a nonlinear transformation of the measurements, which is simple to implement in embedded hardware. A theoretical justification is provided, and the statistical properties of the test statistic are presented in closed form. The method is compared with the standard likelihood ratio test on both simulated data and real data from field tests, where very high detection rates are reported, despite the presence of sensor saturation, measurement noise, and near-field effects of the magnetic field.

This paper presents modeling of wave propagation in pavements from a system identification point of view. First, a model based on the physical structure is derived. Second, experiment design and evaluation are discussed and maximum-likelihood estimators for estimating the model parameters are introduced, assuming an error-in-variables setting. Finally, the proposed methods are applied to measurement data from two experiments under varying environmental conditions. It is found that the proposed methods can be used to estimate the dispersion curves of the considered waveguide and the results can be used for further analysis

In many modern societies, different sensors have started to penetrate life in many new ways.Examples include personal devices for monitoring health and well-being, supervision and power distribution in the smart grid, or smart farming that takesweather and soil conditions into account to efficiently cultivate fields. In order to obtain the desired data, such sensor systems requirewell-designed signal processing algorithms that infer the parameters of interest from the measured quantities. In this thesis, algorithms for traffic monitoring using road side sensors are proposed. Thesensors considered are a combination of an accelerometer measuring road surfacewaves and a magnetometer measuring magnetic disturbances, both caused by vehicles passing the sensors.The research problems addressed are: (1) the feasibility of using road surface waves fortraffic monitoring, (2) the modeling of road surface waves, and (3) combining the measurements of the accelerometer and the magnetometer.These three problems are addressed in the six research papers composing this thesis. First, it is shown that it is indeed viable to exploit road surface waves for estimating vehicle parameters and research challenges are identified by analyzing a first field test. Based on these conclusions, it is shown how to model waves in pavements using system identification and a semi-parametric wave propagation model. Furthermore, an efficient algorithm for estimating the driving direction using magnetometers only is proposed and evaluated. Finally, it is shown how to combine the two sensors. First, an iterative particle smootherbased system identification algorithm is used to jointly estimate the vehicle trajectory as wellas unknown parameters in the system model. Second, a multi-rate particle filter is proposed where unknown parameters are treated through marginalization.Based on the work in this thesis, future research directions are proposed. These include the improvement of some of the models to address problems encountered in the trajectory estimation and tracking algorithms aswell as further development of the estimation methods to make them more efficient and take prior information and constraints into account.

This paper addresses a novel method for vehicle tracking using an extended Kalman filter and measurements of road surface vibrations from a single accelerometer. First, a measurement model for vibrations caused by vehicular road traffic is developed. Then the identifiability of the involved parameters is analyzed. Finally, the measurement model is combined with a constant speed motion model and the Kalman filter is derived. Simulation and measurement results indicate that the approach is feasible and show where further development is needed.