# Publikationer Institutionen för systemteknik

DatorseendeDatorteknik

Elektroniska Kretsar och System

Fordonssystem

Informationskodning

Kommunikationssystem

Reglerteknik

## Senaste doktorsavhandlingarna

Electrification of vehicles is an indispensable step in improving fuel economy and reducing fossil fuel emissions. In particular, hybrid electric vehicle market has gained popularity as one such reliable solution. With the global rise in environmental concerns, the need for advancement of the relevant technologies has become more noticeable than before. In this pursuit, it is well-known that design of effective energy management strategies (EMS) that govern power distribution among the onboard energy sources is key in reducing fuel consumption and its adverse environmental impacts. This thesis is concerned with EMS design for series hybrid electric vehicles from two standpoints.

Powertrain component durability is often neglected in EMS development. In particular, batteries are prone to degradation through usage, a phenomenon widely known as cycle aging, and contribute largely to vehicle cost. In the first part of the thesis, therefore, battery lifetime optimization is integrated into the design of fuel-efficient energy management strategies. An empirical capacity degradation model is adopted from the literature and is modified in order to predict battery lifetime. The multi-objective problem is to compromise between fuel consumption reduction and battery wear minimization. The problem is formulated within two control theory frameworks, namely Pontryagin's minimum principle and model predictive control. Simulation results suggest that there is an enormous potential in prolonging battery lifetime by sacrificing negligible to no excessive amount of fuel consumption. Performance of the developed methodology in the Pontryagin's minimum principle framework exhibits an inverse correlation with the root-mean-square of power request of drive cycles. The results can be used to develop real-time rule-based methods.

The application considered in this part is a hybrid electric wheel loader. While prolonging battery lifetime is economically beneficial for any hybrid electric vehicle, the cost savings for high power applications such as the aforementioned construction equipment can be even more rewarding.

The second part of the thesis is dedicated to the development of time-efficient energy management strategies. Considering the need for real-time feasibility, satisfactory fuel economy and low computation time are the key elements in EMS design. In the first step, the analytical solution to equivalent consumption minimization strategy (ECMS) for series hybrid electric vehicles is derived, where the system constraints are directly taken into account in the derivation process. The equivalence factor bounds are found and used to develop a real time adaptive ECMS. The obtained fuel economy figures are observed to be very close to the non-causal benchmarks. These results are then utilized to propose real-time predictive ECMS algorithms. Two scenarios are investigated depending on the availability of drive cycle knowledge. The first scenario corresponds to vehicles that are expected to follow certain drive cycles. This situation is common among construction machinery such as the wheel loader under study. On the other hand, there are situations where driving mission is not known in advance and the driver behavior is unpredictable, such as typical city driving. For each scenario, an algorithm is presented to compute the equivalence factor efficiently. The control action is then determined by the analytical policy derived previously. Simulations of the developed algorithms on the hybrid wheel loader and a passenger car demonstrate that the methodologies are computationally efficient and attain satisfactory fuel economy with respect to the dynamic programming benchmarks.

```
@phdthesis{diva2:1587506,
author = {Shafikhani, Iman},
title = {{Energy management strategy design for series hybrid electric vehicles}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2165}},
year = {2021},
address = {Sweden},
}
```

Collective decision-making refers to a process in which the agents of a community exchange opinions with the objective of reaching a common decision. It is often assumed that a collective decision is reached through collaboration among the individuals. However in many contexts, concerning for instance collective human behavior, it is more realistic to assume that the agents can collaborate or compete with each other. In this case, different types of collective behavior can be observed. This thesis investigates collective decision-making problems in multiagent systems, both in the case of collaborative and of antagonistic interactions.

The first problem studied in the thesis is a special instance of the consensus problem, denoted "interval consensus" in this work. It consists in letting the agents impose constraints on the possible common consensus value. It is shown that introducing saturated nonlinearities in the decision-making dynamics to describe how the agents express their opinions effectively allows the agents to influence the achievable consensus value and steer it to the intersection of all the intervals imposed by the agents.

A second class of collective decision-making models discussed in the thesis is obtained by replacing the saturations with sigmoidal nonlinearities. This nonlinear interconnected model is first investigated in the collaborative case and then in the antagonistic case, represented as a signed graph of interactions. In both cases, it is shown that the behavior of the model can be described by means of bifurcation analysis, with the equilibria of the system encoding the possible decisions for the community. A scalar positive parameter, denoted "social effort", is added to the model to represent the strength of commitment between the agents, and plays the role of bifurcation parameter in the analysis. It is shown that if the social effort is small, then the community is in a deadlock situation (i.e., no decision is taken), while if the agents have the "right" amount of commitment two alternative consensus decision states for the community are achieved. However, by further increasing the social effort, the agents may fall in a situation of "overcommitment" where multiple (more than 2) decisions are possible. When antagonistic interactions between the agents are taken into account, they may lead to conflicts or social tensions during the decision-making process, which can be quantified by the notion of "frustration" of the signed network representing the community. The aim is to understand how the presence of antagonism (represented by the amount of frustration of the signed network) influences the collective decision-making process. It is shown that, while the qualitative behavior of the system does not change, the value of social effort required from the agents to break the deadlock (i.e., the value for which the bifurcation is crossed) increases with the frustration of the signed network: the higher the frustration, the higher the required social commitment.

A natural context to apply these results is that of political decision-making. In particular it is shown in the thesis how the government formation process in parliamentary democracies can be modeled as a collective decision-making system, where the agents are the parliamentary members, the decision is the vote of confidence they cast to a candidate cabinet coalition, and the social effort parameter is a proxy for the duration of the government negotiation talks. A signed network captures the alliances/rivalries between the political parties in the parliament. The idea is that the frustration of the parliamentary networks should correlate well with the duration of the government negotiation, and it is supported by the analysis of the legislative elections in 29 European countries in the last 40 years.

The final contribution of this thesis is an analysis of the structure of (signed) Laplacian matrices and of their pseudoinverses. It is shown that the pseudoinverse of a Laplacian is in general a signed Laplacian, and in particular that the set of eventually exponentially positive Laplacian matrices (i.e., matrices whose exponential is a matrix with negative entries which becomes and stays positive at a certain power) is closed under stability and matrix pseudoinversion.

```
@phdthesis{diva2:1585664,
author = {Fontan, Angela},
title = {{Collective decision-making on networked systems in presence of antagonistic interactions}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2166}},
year = {2021},
address = {Sweden},
}
```

In computer vision, the aim is to model and extract high-level information from visual sensor measurements such as images, videos and 3D points. Since visual data is often high-dimensional, noisy and irregular, achieving robust data modeling is challenging. This thesis presents works that address challenges within a number of different computer vision problems.

First, the thesis addresses the problem of phase unwrapping for multi-frequency amplitude modulated time-of-flight (ToF) ranging. ToF is used in depth cameras, which have many applications in 3D reconstruction and gesture recognition. While amplitude modulation in time-of-flight ranging can provide accurate measurements for the depth, it also causes depth ambiguities. This thesis presents a method to resolve the ambiguities by estimating the likelihoods of different hypotheses for the depth values. This is achieved by performing kernel density estimation over the hypotheses in a spatial neighborhood of each pixel in the depth image. The depth hypothesis with the highest estimated likelihood can then be selected as the output depth. This approach yields improvements in the quality of the depth images and extends the effective range in both indoor and outdoor environments.

Next, point set registration is investigated, which is the problem of aligning point sets from overlapping depth images or 3D models. Robust registration is fundamental to many vision tasks, such as multi-view 3D reconstruction and object pose estimation for robotics. The thesis presents a method for handling density variations in the measured point sets. This is achieved by modeling a latent distribution representing the underlying structure of the scene. Both the model of the scene and the registration parameters are inferred in an Expectation-Maximization based framework. Secondly, the thesis introduces a method for integrating features from deep neural networks into the registration model. It is shown that the deep features improve registration performance in terms of accuracy and robustness. Additionally, improved feature representations are generated by training the deep neural network end-to-end by minimizing registration errors produced by our registration model.

Further, an approach for 3D point set segmentation is presented. As scene models are often represented using 3D point measurements, segmentation of these is important for general scene understanding. Learning models for segmentation requires a significant amount of annotated data, which is expensive and time-consuming to acquire. The approach presented in the thesis circumvents this by projecting the points into virtual camera views and render 2D images. The method can then exploit accurate convolutional neural networks for image segmentation and map the segmentation predictions back to the 3D points. This also allows for transferring learning using available annotated image data, thereby reducing the need for 3D annotations.

Finally, the thesis explores the problem of video object segmentation (VOS), where the task is to track and segment target objects in each frame of a video sequence. Accurate VOS requires a robust model of the target that can adapt to different scenarios and objects. This needs to be achieved using only a single labeled reference frame as training data for each video sequence. To address the challenges in VOS, the thesis introduces a parametric target model, optimized to predict a target label derived from the mask annotation. The target model is integrated into a deep neural network, where its predictions guide a decoder module to produce target segmentation masks. The deep network is trained on labeled video data to output accurate segmentation masks for each frame. Further, it is shown that by training the entire network model in an end-to-end manner, it can learn a representation of the target that provides increased segmentation accuracy.

```
@phdthesis{diva2:1559711,
author = {Järemo Lawin, Felix},
title = {{Learning Representations for Segmentation and Registration}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2151}},
year = {2021},
address = {Sweden},
}
```

Reducing the fuel consumption of today's vehicle fleet is of great importance due to the environmental impact of using fossil-based fuels. The turbocharged compression ignition (CI) engine is widely used for trucks. The CI engine efficiency is dependent on the operating point, in terms of rotational speed and load. The selection of load point can be controlled by selecting suitable gears, but remains a challenging task during dynamic driving, due to the turbocharger dynamics which introduces a lag in the system. Electric turbocharger technologies can improve the engine response time, but developing efficient control strategies can be challenging. Due to turbocharger lag, all conditions that are reachable in stationary operation for the turbocharged CI engine are not always reachable during dynamic events, for example after an up-shift where the engine speed and torque demand changes rapidly.

In this work the fuel saving potential of electric turbocharging for a heavy-duty truck performing a long-haulage driving mission is investigated. An electric turbocharger control strategy is proposed and evaluated. The results show that the fuel consumption can be reduced using the electric turbocharger, when comparing to a conventional turbocharged CI truck performing a long-haulage driving mission.

A turbocharged CI engine model suitable for optimal control of transient behavior is developed. Sub-models are validated using data describing the components, and the model suitability for optimal control is shown with a tip-in example. To increase the model accuracy, the torque model is extended with a further dependence on the air-fuel ratio and operating point dependent losses. The complete engine model is parameterized for a set of stationary load points. The model is validated using data from a dynamic engine test, where it is shown that both the stationary and dynamic features in the data is represented well by the model. The developed engine model is used as a foundation in an optimal control problem setup to solve fuel optimal accelerations including gear changes. The setup is used to investigate the impact of driveshaft flexibility on the optimal control results, when compared to a stiff driveshaft model. Apart from a slight increase in fuel consumption, the driveshaft flexibility is shown to have minor effects on the fuel optimal control signals, in terms of general torque output and gear shift characteristics.

The hybrid electric vehicle (HEV) technology can potentially reduce the consumption of diesel fuel, but how to design and control the system, consisting of several degrees of freedom remains a challenging task. Energy optimal accelerations of a CI parallel HEV with electric turbocharger is investigated using the optimal control problem setup. The results show that the electric turbocharger is used when the electrical energy cost is high, and the usage of the crank shaft motor is increasing with decreasing electric energy cost.

To summarize, the developed models and problem setups enable investigations of different powertrain configurations and optimal control of these. One conclusion is that the energy savings using an electric turbocharger and crank shaft motor during accelerations are significant.

```
@phdthesis{diva2:1552726,
author = {Ekberg, Kristoffer},
title = {{Modeling and Optimal Control for Dynamic Driving of Hybridized Vehicles with Turbocharged Diesel Engines}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2145}},
year = {2021},
address = {Sweden},
}
```

Cellular network operators have witnessed significant growth in data traffic in the past few decades. This growth occurs due to the increase in the number of connected mobile devices, and further, the emerging mobile applications developed for rendering video-based on-demand services. As the available frequency bandwidth for cellular communication is limited, significant efforts are dedicated to improving the utilization of available spectrum and increasing the system performance with the aid of new technologies. Third-generation (3G) and fourth-generation (4G) mobile communication networks were designed to facilitate high data traffic in cellular networks in past decades. Nevertheless, there is still a requirement for new cellular network technologies to accommodate the ever-growing data traffic demand. The fifth-generation (5G) is the latest generation of mobile communication systems deployed and implemented around the world. Its objective is to meet the tremendous ongoing increase in the data traffic requirements in cellular networks.

Massive MIMO (multiple-input-multi-output) is one of the backbone technologies in 5G networks. Massive MIMO originated from the concept of multi-user MIMO. It consists of base stations (BSs) implemented with a large number of antennas to increase the signal strengths via adaptive beamforming and concurrently serving many users on the same time-frequency blocks. With Massive MIMO technology, there is a notable enhancement of both sum spectral efficiency (SE) and energy efficiency (EE) in comparison with conventional MIMO-based cellular networks. Resource allocation is an imperative factor to exploit the specified gains of Massive MIMO. It corresponds to efficiently allocating resources in the time, frequency, space, and power domains for cellular communication. Power control is one of the resource allocation methods of Massive MIMO networks to deliver high spectral and energy efficiency. Power control refers to a scheme that allocates transmit powers to the data transmitters such that the system maximizes some desirable performance metric.

The first part of this thesis investigates reusing a Massive MIMO network's resources for direct communication of some specific user pairs known as device-to-device (D2D) underlay communication. D2D underlay can conceivably increase the SE of traditional Massive MIMO networks by enabling more simultaneous transmissions on the same frequencies. Nevertheless, it adds additional mutual interference to the network. Consequently, power control is even more essential in this scenario than the conventional Massive MIMO networks to limit the interference caused by the cellular network and the D2D communication to enable their coexistence. We propose a novel pilot transmission scheme for D2D users to limit the interference on the channel estimation phase of cellular users compared with sharing pilot sequences for cellular and D2D users. We also introduce a novel pilot and data power control scheme for D2D underlaid Massive MIMO networks. This method aims to assure that the D2D communication enhances the SE of the network compared to conventional Massive MIMO networks.

In the second part of this thesis, we propose a novel power control approach for multi-cell Massive MIMO networks. The proposed power control approach solves the scalability issue of two well-known power control schemes frequently used in the Massive MIMO literature, based on the network-wide max-min and proportional fairness performance metrics. We first identify the scalability issue of these existing approaches. Additionally, we provide mathematical proof for the scalability of our proposed method. Our scheme aims at maximizing the geometric mean of the per-cell max-min SE. To solve the optimization problem, we prove that it can be rewritten in a convex form and is solved using standard optimization solvers.

The final part of this thesis focuses on downlink channel estimation in a Massive MIMO network. In Massive MIMO networks, to fully benefit from large antennas at the BSs and perform resource allocation, the BS must have access to high-quality channel estimates that can be acquired via the uplink pilot transmission phase. Time-division duplex (TDD) based Massive MIMO relies on channel reciprocity for the downlink transmission. Thanks to the channel hardening in the Massive MIMO networks with ideal propagation conditions, users rely on the statistical knowledge of channels for decoding the data in the downlink. However, when the channel hardening level is low, using only the channel statistics causes fluctuations in the performance. We investigate how to improve the performance by empowering the user to estimate the downlink channel from downlink data transmissions utilizing a model-based and a data-driven approach instead of relying only on channel statistics. Furthermore, the performance of the proposed method is compared with solely relying on statistical knowledge.

```
@phdthesis{diva2:1556280,
author = {Ghazanfari, Amin},
title = {{Multi-Cell Massive MIMO: Power Control and Channel Estimation}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2142}},
year = {2021},
address = {Sweden},
}
```

Early computer vision algorithms operated on dense 2D images captured using conventional monocular or color sensors. Those sensors embrace a passive nature providing limited scene representations based on light reflux, and are only able to operate under adequate lighting conditions. These limitations hindered the development of many computer vision algorithms that require some knowledge of the scene structure under varying conditions. The emergence of active sensors such as Time-of-Flight (ToF) cameras contributed to mitigating these limitations; however, they gave a rise to many novel challenges, such as data sparsity that stems from multi-path interference, and occlusion.

Many approaches have been proposed to alleviate these challenges by enhancing the acquisition process of ToF cameras or by post-processing their output. Nonetheless, these approaches are sensor and model specific, requiring an individual tuning for each sensor. Alternatively, learning-based approaches, i.e., machine learning, are an attractive solution to these problems by learning a mapping from the original sensor output to a refined version of it. Convolutional Neural Networks (CNNs) are one example of powerful machine learning approaches and they have demonstrated a remarkable success on many computer vision tasks. Unfortunately, CNNs naturally operate on dense data and cannot efficiently handle sparse data from ToF sensors.

In this thesis, we propose a novel variation of CNNs denoted as the Normalized Convolutional Neural Networks that can directly handle sparse data very efficiently. First, we formulate a differentiable normalized convolution layer that takes in sparse data and a confidence map as input. The confidence map provides information about valid and missing pixels to the normalized convolution layer, where the missing values are interpolated from their valid vicinity. Afterwards, we propose a confidence propagation criterion that allows building cascades of normalized convolution layers similar to the standard CNNs. We evaluated our approach on the task of unguided scene depth completion and achieved state-of-the-art results using an exceptionally small network.

As a second contribution, we investigated the fusion of a normalized convolution network with standard CNNs employing RGB images. We study different fusion schemes, and we provide a thorough analysis for different components of the network. By employing our best fusion strategy, we achieve state-of-the-art results on guided depth completion using a remarkably small network.

Thirdly, to provide a statistical interpretation for confidences, we derive a probabilistic framework for the normalized convolutional neural networks. This framework estimates the input confidence in a self-supervised manner and propagates it to provide a statistically valid output confidence. When compared against existing approaches for uncertainty estimation in CNNs such as Bayesian Deep Learning, our probabilistic framework provides a higher quality measure of uncertainty at a significantly lower computational cost.

Finally, we attempt to employ our framework in a common task in CNNs, namely upsampling. We formulate the upsampling problem as a sparse problem, and we employ the normalized convolutional neural networks to solve it. In comparison to existing approaches, our proposed upsampler is structure-aware while being light-weight. We test our upsampler with various optical flow estimation networks, and we show that it consistently improves the results. When integrated with a recent optical flow network, it sets a new state-of-the-art on the most challenging optical flow dataset.

```
@phdthesis{diva2:1547851,
author = {Eldesokey, Abdelrahman},
title = {{Uncertainty-Aware Convolutional Neural Networks for Vision Tasks on Sparse Data}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2123}},
year = {2021},
address = {Sweden},
}
```

In less than ten years, deep neural networks have evolved into all-encompassing tools in multiple areas of science and engineering, due to their almost unreasonable effectiveness in modeling complex real-world relationships. In computer vision in particular, they have taken tasks such as object recognition, that were previously considered very difficult, and transformed them into everyday practical tools. However, neural networks have to be trained with supercomputers on massive datasets for hours or days, and this limits their ability adjust to changing conditions.

This thesis explores discriminative correlation filters, originally intended for tracking large objects in video, so-called visual object tracking. Unlike neural networks, these filters are small and can be quickly adapted to changes, with minimal data and computing power. At the same time, they can take advantage of the computing infrastructure developed for neural networks and operate within them.

The main contributions in this thesis demonstrate the versatility and adaptability of correlation filters for various problems, while complementing the capabilities of deep neural networks. In the first problem, it is shown that when adopted to track small regions and points, they outperform the widely used Lucas-Kanade method, both in terms of robustness and precision.

In the second problem, the correlation filters take on a completely new task. Here, they are used to tell different places apart, in a 16 by 16 square kilometer region of ocean near land. Given only a horizon profile - the coast line silhouette of islands and islets as seen from an ocean vessel - it is demonstrated that discriminative correlation filters can effectively distinguish between locations.

In the third problem, it is shown how correlation filters can be applied to video object segmentation. This is the task of classifying individual pixels as belonging either to a target or the background, given a segmentation mask provided with the first video frame as the only guidance. It is also shown that discriminative correlation filters and deep neural networks complement each other; where the neural network processes the input video in a content-agnostic way, the filters adapt to specific target objects. The joint function is a real-time video object segmentation method.

Finally, the segmentation method is extended beyond binary target/background classification to additionally consider distracting objects. This addresses the fundamental difficulty of coping with objects of similar appearance.

```
@phdthesis{diva2:1545394,
author = {Robinson, Andreas},
title = {{Discriminative correlation filters in robot vision}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2146}},
year = {2021},
address = {Sweden},
}
```

In the last decade, developments in hardware, sensors and software have made it possible to create increasingly autonomous systems. These systems can be as simple as limited driver assistance software lane-following in cars, or limited collision warning systems for otherwise manually piloted drones. On the other end of the spectrum exist fully autonomous cars, boats or helicopters. With increasing abilities to function autonomously, the demands to operate with minimal human supervision in unstructured environments increase accordingly.

Common to most, if not all, autonomous systems is that they require an accurate model of the surrounding world. While there is currently a large number of possible sensors useful to create such models available, cameras are one of the most versatile. From a sensing perspective cameras have several advantages over other sensors in that they require no external infrastructure, are relatively cheap and can be used to extract such information as the relative positions of other objects, their movements over time, create accurate maps and locate the autonomous system within these maps.

Using cameras to produce a model of the surroundings require solving a number of technical problems. Often these problems have a basis in recognizing that an object or region of interest is the same over time or in novel viewpoints. In visual tracking this type of recognition is required to follow an object of interest through a sequence of images. In geometric problems it is often a requirement to recognize corresponding image regions in order to perform 3D reconstruction or localization.

The first set of contributions in this thesis is related to the improvement of a class of on-line learned visual object trackers based on discriminative correlation filters. In visual tracking estimation of the objects size is important for reliable tracking, the first contribution in this part of the thesis investigates this problem. The performance of discriminative correlation filters is highly dependent on what feature representation is used by the filter. The second tracking contribution investigates the performance impact of different features derived from a deep neural network.

A second set of contributions relate to the evaluation of visual object trackers. The first of these are the visual object tracking challenge. This challenge is a yearly comparison of state-of-the art visual tracking algorithms. A second contribution is an investigation into the possible issues when using bounding-box representations for ground-truth data.

In real world settings tracking typically occur over longer time sequences than is common in benchmarking datasets. In such settings it is common that the model updates of many tracking algorithms cause the tracker to fail silently. For this reason it is important to have an estimate of the trackers performance even in cases when no ground-truth annotations exist. The first of the final three contributions investigates this problem in a robotics setting, by fusing information from a pre-trained object detector in a state-estimation framework. An additional contribution describes how to dynamically re-weight the data used for the appearance model of a tracker. A final contribution investigates how to obtain an estimate of how certain detections are in a setting where geometrical limitations can be imposed on the search region. The proposed solution learns to accurately predict stereo disparities along with accurate assessments of each predictions certainty.

```
@phdthesis{diva2:1545918,
author = {Häger, Gustav},
title = {{Learning visual perception for autonomous systems}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2138}},
year = {2021},
address = {Sweden},
}
```

Many practical applications, such as search and rescue operations and environmental monitoring, involve the use of mobile sensor platforms. The workload of the sensor operators is becoming overwhelming, as both the number of sensors and their complexity are increasing. This thesis addresses the problem of automating sensor systems to support the operators. This is often referred to as sensor management. By planning trajectories for the sensor platforms and exploiting sensor characteristics, the accuracy of the resulting state estimates can be improved. The considered sensor management problems are formulated in the framework of stochastic optimal control, where prior knowledge, sensor models, and environment models can be incorporated. The core challenge lies in making decisions based on the predicted utility of future measurements.

In the special case of linear Gaussian measurement and motion models, the estimation performance is independent of the actual measurements. This reduces the problem of computing sensing trajectories to a deterministic optimal control problem, for which standard numerical optimization techniques can be applied. A theorem is formulated that makes it possible to reformulate a class of nonconvex optimization problems with matrix-valued variables as convex optimization problems. This theorem is then used to prove that globally optimal sensing trajectories can be computed using off-the-shelf optimization tools.

As in many other fields, nonlinearities make sensor management problems more complicated. Two approaches are derived to handle the randomness inherent in the nonlinear problem of tracking a maneuvering target using a mobile range-bearing sensor with limited field of view. The first approach uses deterministic sampling to predict several candidates of future target trajectories that are taken into account when planning the sensing trajectory. This significantly increases the tracking performance compared to a conventional approach that neglects the uncertainty in the future target trajectory. The second approach is a method to find the optimal range between the sensor and the target. Given the size of the sensor's field of view and an assumption of the maximum acceleration of the target, the optimal range is determined as the one that minimizes the tracking error while satisfying a user-defined constraint on the probability of losing track of the target.

While optimization for tracking of a single target may be difficult, planning for jointly maintaining track of discovered targets and searching for yet undetected targets is even more challenging. Conventional approaches are typically based on a traditional tracking method with separate handling of undetected targets. Here, it is shown that the Poisson multi-Bernoulli mixture (PMBM) filter provides a theoretical foundation for a unified search and track method, as it not only provides state estimates of discovered targets, but also maintains an explicit representation of where undetected targets may be located. Furthermore, in an effort to decrease the computational complexity, a version of the PMBM filter which uses a grid-based intensity to represent undetected targets is derived.

```
@phdthesis{diva2:1541009,
author = {Boström-Rost, Per},
title = {{Sensor Management for Target Tracking Applications}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2137}},
year = {2021},
address = {Sweden},
}
```

During the last decades, motion planning for autonomous systems has become an important area of research. The high interest is not the least due to the development of systems such as self-driving cars, unmanned aerial vehicles and robotic manipulators. The objective in optimal motion planning problems is to find feasible motion plans that also optimize a performance measure. From a control perspective, the problem is an instance of an optimal control problem. This thesis addresses optimal motion planning problems for complex dynamical systems that operate in unstructured environments, where no prior reference such as road-lane information is available. Some example scenarios are autonomous docking of vessels in harbors and autonomous parking of self-driving tractor-trailer vehicles at loading sites. The focus is to develop optimal motion planning algorithms that can reliably be applied to these types of problems. This is achieved by combining recent ideas from automatic control, numerical optimization and robotics.

The first contribution is a systematic approach for computing local solutions to motion planning problems in challenging unstructured environments. The solutions are computed by combining homotopy methods and direct optimal control techniques. The general principle is to define a homotopy that transforms, or preferably relaxes, the original problem to an easily solved problem. The approach is demonstrated in motion planning problems in 2D and 3D environments, where the presented method outperforms a state-of-the-art asymptotically optimal motion planner based on random sampling.

The second contribution is an optimization-based framework for automatic generation of motion primitives for lattice-based motion planners. Given a family of systems, the user only needs to specify which principle types of motions that are relevant for the considered system family. Based on the selected principle motions and a selected system instance, the framework computes a library of motion primitives by simultaneously optimizing the motions and the terminal states.

The final contribution of this thesis is a motion planning framework that combines the strengths of sampling-based planners with direct optimal control in a novel way. The sampling-based planner is applied to the problem in a first step using a discretized search space, where the system dynamics and objective function are chosen to coincide with those used in a second step based on optimal control. This combination ensures that the sampling-based motion planner provides a feasible motion plan which is highly suitable as warm-start to the optimal control step. Furthermore, the second step is modified such that it also can be applied in a receding-horizon fashion, where the proposed combination of methods is used to provide theoretical guarantees in terms of recursive feasibility, worst-case objective function value and convergence to the terminal state. The proposed motion planning framework is successfully applied to several problems in challenging unstructured environments for tractor-trailer vehicles. The framework is also applied and tailored for maritime navigation for vessels in archipelagos and harbors, where it is able to compute energy-efficient trajectories which complies with the international regulations for preventing collisions at sea.

```
@phdthesis{diva2:1537293,
author = {Bergman, Kristoffer},
title = {{Exploiting Direct Optimal Control for Motion Planning in Unstructured Environments}},
school = {Linköping University},
type = {{Linköping Studies in Science and Technology. Dissertations No. 2133}},
year = {2021},
address = {Sweden},
}
```

Senast uppdaterad: 2020-10-01