Autonomous driving requires tactical decision-making while navigating in a dynamic shared space environment. The complexity and uncertainty in this process arise due to unknown and tightly-coupled interaction among traffic users. This thesis work formulates an unknown navigation problem as a Markov decision process (MDP), supported by models of traffic participants and userspace. Instead of modeling a traditional MDP, this work formulates a Multi-policy decision making (MPDM) in a shared space scenario with pedestrians and vehicles. The employed model enables a unified and robust self-driving of the ego vehicle by selecting a desired policy along the pre-planned path. Obstacle avoidance is coupled within the navigation module performing a detour off the planned path and obtaining a reward on task completion and penalizing for collision with others. In addition to this, the thesis work is further extended by analyzing the real-time constraints of the proposed model. The performance of the implemented framework is evaluated in a simulation environment on a typical construction (quarry) scenario. The effectiveness and efficiency of the elected policy verify the desired behavior of the autonomous vehicle.

Den ökade datakraften och digitaliseringen av arbete har bidragit till möjligheten att automatisera processer. En metod för att automatisera processer är med Robotic Process Automation (RPA), vilket utför regelstyrda uppgifter baserat på detaljerade processbeskrivningar. Användningen av RPA blir allt vanligare inom den offentliga sektorn. Ett exempel på detta är Trafikverket som infört RPA för att automatisera processer med målet att frigöra tid för medarbetare. Denna studie har undersökt vad Trafikverkets införande och användning av RPA har inneburit för den digitala arbetsmiljön. Studien har undersökt detta genom att genomföra intervjuer med medarbetare som påverkats till följd av införandet och användning av RPA eller varit delaktig i införandet av RPA-lösningen. Studiens bidrag skapar en ökad förståelse för vad införandet och användandet av RPA har för konsekvenser på digital arbetsmiljö i organisationer i den offentliga sektorn genom att presentera tre konsekvenser. Studien visar att användningen av RPA kan resultera i en initial oro och en upplevd minskad anställningstrygghet bland medarbetarnas. Nya arbetsuppgifter kan uppstå till följd av införandet av RPA, dessa kan upplevas som monotona och irritationsmoment av medarbetarna. Slutligen kan det innebära att medarbetare får mer tid till varierande och kvalificerade arbetsuppgifter med möjlighet till kompetensutveckling. Detta resulterar i att medarbetarna uttrycker en ökad förändringsbenägenhet gentemot fortsatt automatisering med RPA.

Voice is the primary tool of communication to a majority of people on earth. Humans are wired to process speech, meaning voice interaction require little cognitive effort. Advancements in voice technology over the last 20 years have seen an increased prevalence of voice-controlled applications. However, false expectations can potentially cause severe interaction deficiencies to many of these voice user interfaces. One of the many application areas being connected to voice is access solutions such as smart locks. With a fundamental value proposition of keeping people and their belongings safe and secure in convenient ways, access solutions require that many of the design decisions are delicate. The present thesis aimed to evaluate how expectations of access solutions in home environments can be affected by different voice attributes. A literature study was conducted to explore the rich body of research on the topic of voice technology and the psychological effects of synthesized speech. Based on the literature study, a design process with recognized methods for developing voice user interfaces was conducted. The design process led up to a Wizard of Oz test that was used to evaluate how different conversational strategies and voices affected expectations and perception of a voice-controlled smart lock. The results showed that choosing an appropriate conversational style is fundamental to provide users with a sense of control. Furthermore, the study provided insight on how previous experience of interacting with voice-controlled devices have an impact on the feeling of personalization in gendered synthesized voices. Finally, the study discusses some ethical considerations that have to be made when designing voice user interfaces that ultimately should provide value to users, not confine their privacy. / Röst är den primära kommunikationskällan för de flesta människorna på jorden. Förmågan att bearbeta tal är något människor föds med, vilket gör att interaktion genom röst kräver liten kognitiv ansträngning. Framsteg inom röstteknologi under de senaste 20 åren har lett till ett ökat utbud av röststyrda applikationer. För många av dessa röststyrda applikationer existerar en risk att falska förväntningar leder till avsevärt försämrad interaktion. Ett av många applikationsområden där röst börjar framträda är accesslösningar såsom smarta lås. Med sitt fundamentala värde att hålla människor och deras tillgångar trygga och säkra utan att kompromissa enkelheten, kräver utformningen av accesslösningar flera delikata beslut. Därmed har studien ämnat att utvärdera hur användares förväntningar på röststyrda accesslösningar påverkas av röstattribut. För att undersöka och öka förståelsen av röstteknologi och de psykologiska effekterna av tal, genomfördes en litteraturstudie av existerande forskning initialt. Baserat på fynden i litteraturstudien startades sedan en designprocess för utveckling av ett röstgränssnitt. Erkända metoder användes för framtagandet av ett användarvänligt gränssnitt. Designprocessen låg till grund för ett Wizard of Oz test där olika konversationsstrategier och röster påverkade förväntningar av och uppfattningen av ett röststyrt smart lås. Resultaten visar att det är fundamentalt att välja en passande konversationsstil för att ge användare en känsla av kontroll. Studien påvisade även hur tidigare erfarenheter av röststyrda applikationer påverkar förväntningar av andra röststyrda applikationer och gör dem mer eller mindre personliga. Vidare diskuteras etiska avvägningar som måste göras när man designar röstgränssnitt som ska medföra användarvärde och inte inskränka användarnas integritet.

With the simulations tools available today, simulation can be utilised as a platform for more advanced software testing. By introducing simulations to software testing of robot controllers, the motion performance testing phase can begin at an earlier stage of development. This would benefit all parties involved with the robot controller. Testers at ABB would be able to include more motion performance tests to the regression tests. Also, ABB could save money by adapting to simulated robot tests and customers would be provided with more reliable software updates. In this thesis, a method is developed utilising simulations to create a test set for detecting motion anomalies in new robot controller versions. With auto-generated test cases and a similarity analysis that calculates the Hausdorff distance for a test case executed on controller versions with an induced artificial bug. A test set has been created with the ability to detect anomalies in a robot controller with a bug.

Autonomous driving is increasingly popular among people and automotive industries in realizing their presence both in passenger and goods transportation. Safer autonomous navigation might be very challenging if there is a failure in sensing system. Among several sensing systems, image classification plays a major role in understanding the road signs and to regulate the vehicle control based on urban road rules. Hence, a robust classifier algorithm irrespective of camera position, view angles, environmental condition, different vehicle size & type (Car, Bus, Truck, etc.,) of an autonomous platform is of prime importance. In this study, Convolutional Neural Network (CNN) based classifier algorithm has been implemented to ensure improved robustness for recognizing traffic signs. As training data play a crucial role in supervised learning algorithms, there come an effective dataset requirement which can handle dynamic environmental conditions and other variations caused due to the vehicle motion (will be referred as challenges). Since the collected training data might not contain all the dynamic variations, the model weakness can be identified by exposing it to variations (Blur, Darkness, Shadow, etc.,) faced by the vehicles in real-time as a initial testing sequence. To overcome the weakness caused due to the training data itself, an effective augmentation technique enriching the training data in order to increase the model capacity for withstanding the variations prevalent in urban environment has been proposed. As a major contribution, a framework has been developed to identify model weakness and successively introduce a targeted augmentation methodology for classification improvement. Targeted augmentation is based on estimated weakness caused due to the challenges with difficulty levels, only those necessary for better classification were then augmented further. Predictive Augmentation (PA) and Predictive Multiple Augmentation (PMA) are the two proposed methods to adapt the model based on targeted challenges by delivering with high numerical value of confidence. We validated our framework on two different training datasets (German Traffic Sign Recognition Benchmark (GTSRB) and Heavy Vehicle data collected from bus) and with 5 generated test groups containing varying levels of challenge (simple to extreme). The results show impressive improvement by ≈ 5-20% in overall classification accuracy thereby keeping their high confidence.

Emulation is a technology, historically mostly used for virtual commissioning of automated industrial systems, and operator training. Trends show that new areas for deployment are being investigated. One way to broaden the scope of emulation technology is to increase emulation detail level. The University of Skövde conduct research within emulation technology, and are developing a higher detail level emulation platform performing  on component level. For transparent and systematic development of component models on this level, an open, extensible, and flexible data model for emulation models of industrial components is wanted. This thesis is contributing to this endeavour by developing a first draft of such a data model. A demonstration is also conducted by implementing a few components into the developing emulation environment, using XML as file format. An iterative "design and creation" methodology was used to develop and implement an object oriented data model. A selected set of industrial components were used to develop and demonstrate the data model, and the final result is visually represented as a class diagram together with explanatory documentation. Using the methodology and data modelling strategy used in this thesis, systematic and transparent development of emulation models on component level is possible in an extensible and flexible manner. / Emulering är en teknologi som historiskt mestadels använts vid virtuel idrifttagning av industriella automatiserade system samt vid operatörsträning. Trender visar att nya användningsområden utforskas. Ett sätt att vidga användningsområdet för emulering är att öka dess detaljnivå. Högskolan i Skövde utför forskning inom emulering och utvecklar en emuleringsplattform med utökad detaljnivå, även kallad komponentnivån. För att kunna arbeta systematiskt med utvecklandet av emuleringsmodeller för denna nivå önskas en öppen, skalbar, och flexibel datamodell för emuleringsmodeller. Detta examensarbete bidrar till detta genom att utveckla ett första utkast av en sådan data modell. Datamodellen demonstreras genom implementation inom den utvecklandes emuleringsmiljön, med hjälp av filformatet XML. En iterativ "design and creation" metodologi användes för att utveckla och implementera datamodellen. Ett set av industriella komponenter användes i utvecklingen och implementationen av datamodellen. Projektets resultat presenteras som ett klassdiagram tillsammans med förklarande dokumentation. Används projektes metodologi och datamodellerings-strategi kan man med fördel arbeta transparant och systematiskt med utveckling av emuleringsmodeller för anginven nivå. / TWIN

A growing population is driving automatization in agricultural industry to strive for more productive arable land. Being part of this process, this work is aimed to investigate the possibility to implement sensor-based automation in a particular system called Front End Loader, which is a lifting arms that is commonly mounted on the front of a tractor. Two main tasks are considered here, namely Electronic Self Leveling (ESL) and payload estimation. To propose commercially implementable solutions for these tasks, specific objectives are set, which are: 1) to propose a controller to perform ESL under typical disturbances 2) to propose a methodology for payload estimation considering realistic estimation conditions. Lastly, aligned with these goals, 3) to propose models for the Front End Loader under consideration for derivation of solutions of the specified tasks. The self-leveling task assists farmers in maintaining the angular position of the mounted implements, e.g. a bale handler or a bucket, with respect to the ground when the loader is manually lifted or lowered. Experimental results show that different controllers are required in lifting and lowering motions to maintain the implement's angular position with a required accuracy due to principle differences in gravity impact. The gravity helps the necessary correction in lifting motion, but works against the correction in lowering motions. This led us to propose a controller with a proportional term, a discontinuous term and an on-line disturbance estimation and compensation as well as the tuning procedure to achieve a 2 degrees tracking error for lowering motions in steady state. The proposed controller shows less sensitive performance to lowering velocity, as the main disturbance, in comparison to a linear controller. The second task, payload estimation, assists farmers to work within safety range as well as to work with a weight measurement tool. A mechanical model derived based on equations of motion is improved by a pressure based friction to sufficiently accurately represent the motion of the front end loader under consideration. The proposed model satisfies the desired estimation accuracy of 2\% full scale error in a certain estimation condition domain in constant velocity regions, with off-line calibration step and off-line payload estimation step. An on-line version of the estimation based on Recursive Least Squares also fulfills the desired accuracy, while keeping the calibration step off-line.

This thesis deals with control aspects of complex hydromechanical transmissions. The overall purpose is to increase the knowledge of important aspects to consider during the development of hydromechanical transmissions to ensure transmission functionality. These include ways of evaluating control strategies in early design stages as well as dynamic properties and control aspects of displacement controllers, which are key components in these systems. Fuel prices and environmental concerns are factors that drive research on propulsion in heavy construction machinery. Hydromechanical transmissions are strong competitors to conventional torque-converter transmissions used in this application today. They offer high efficiency and wide speed/torque conversion ranges, and may easily be converted to hybrids that allow further fuel savings through energy recuperation. One challenge with hydromechanical transmissions is that they offer many different configurations, which in turn makes it important to enable evaluation of control aspects in early design stages. In this thesis, hardware-in-the-loop simulations, which blend hardware tests and standard software-based simulations, are considered to be a suitable method. A multiple-mode transmission applied to a mid-sized construction machine is modelled and evaluated in offline simulations as well as in hardware-in-the-loopsimulations. Hydromechanical transmissions rely on efficient variable pumps/motors with fast, accurate displacement controllers. This thesis studies the dynamic behaviour of the displacement controller in swash-plate axial-piston pumps/motors. A novel control approach in which the displacement is measured with an external sensor is proposed. Performance and limitations of the approach are tested in simulations and in experiments. The experiments showed a significantly improved performance with a controller that is slightly more advanced than a standard proportional controller. The implementation of the controller allows simple tuning and good predictability of the displacement response.

Ego-motion sensors are commonly used for pose estimation in Simultaneous Localization And Mapping (SLAM) algorithms. Inertial Measurement Units (IMUs) are popular sensors but suffer from integration drift over longer time scales. To remedy the drift they are often used in combination with additional sensors, such as a LiDAR. Pose estimation is used when scans, produced by these additional sensors, are being matched. The matching of scans can be computationally heavy as one scan can contain millions of data points. Methods exist to simplify the problem of finding the relative pose between sensor data, such as the Normal Distribution Transform SLAM algorithm. The algorithm separates the point cloud data into a voxelgrid and represent each voxel as a normal distribution, effectively decreasing the amount of data points. Registration is based on a function which converges to a minimum. Sub-optimal conditions can cause the function to converge at a local minimum. To remedy this problem this thesis explores the benefits of combining IMU sensor data to estimate the pose to be used in the NDT SLAM algorithm.

It is not reliable to depend on a persons inference on dense data of high dimensionality on a daily basis. A person will grow tired or become distracted and make mistakes over time. Therefore it is desirable to study the feasibility of replacing a persons inference with that of Machine Learning in order to improve reliability. One-Class Support Vector Machines (SVM) with three different kernels (linear, Gaussian and polynomial) are implemented and tested for Anomaly Detection. Principal Component Analysis is used for dimensionality reduction and autoencoders are used with the intention to increase performance. Standard soft-margin SVMs were used for multi-class classification by utilizing the 1vsAll and 1vs1 approaches with the same kernels as for the one-class SVMs. The results for the one-class SVMs and the multi-class SVM methods are compared against each other within their respective applications but also against the performance of Back-Propagation Neural Networks of varying sizes. One-Class SVMs proved very effective in detecting anomalous samples once both Principal Component Analysis and autoencoders had been applied. Standard SVMs with Principal Component Analysis produced promising classification results. Twin SVMs were researched as an alternative to standard SVMs.