<p>Denna rapport behandlar en fullständig metod för att följa, associera och klassificera ekospår. Vid radarnivåmätning ska ekon som härstammar ifrån samma fysikaliska objekt följas över tiden och klassificeras. Dagens radarnivågivare löser problemet genom konfiguration. Tankens geometri samt vilka störande objekt som finns anges vid installation.</p><p>Den framtagna metoden har delats upp i tre delmoment. Först binds individuella inmätta ekon samman till linjesegment. Dessa segment associeras samman till ekospår varefter dessa klassas fysikaliskt. Metoden kräver betydligt mindre konfiguration än den lösning som används för tillfället.</p> / <p>In this report a method for target tracking, association and classification of echo tracks in radar level gauges is discussed. In tank gauging echoes from the same physical object are called an echo track. The echoes must be tracked over time and classified. Currently radar level gauge units requires configuration to achieve classifications of echo</p><p>tracks. At installation tank geometry and disturbing echoes has to be registered.</p><p>The method has been divided into three parts. Firstly individual measured echoes are replaced by line segments. Association algorithms then create echo tracks from these segments. The third step is to classify these tracks physically. This method eliminates registering of disturbing echoes.</p>

For the environment of target tracking with highly non-linear models and non-Gaussian noise, the tracking performance of the particle filter is better than extended Kalman filter; in addition, the design of particle filter is simpler, so it is quite suitable for the realistic environment. However, particle filter depends on the probability model of the noise. If the knowledge of the noise is incorrect, the tracking performance of the particle filter will degrade severely. To tackle the problem, cost function-based particle filters have been studied. Though suffering from minor degradation on the performance, the cost function-based particle filters do not need probability assumptions of the noises. The application of cost function-based particle filters will be more robust in any realistic environment. Cost function-based particle filters will enable maneuvering multiple target tracking to be suitable for any environment because it does not depend on the noise model. The difficulty lies in the link between the estimator and data association. The likelihood function are generally obtained from the algorithm of the data association; while cost functions are used in the cost function-based particle filter for moving the particles and update the corresponding weights without probability assumptions on the noises. The thesis is focused on the combination of data association and cost function-based particle filter, in order to make the algorithm of multiple target tracking more robust in noisy environments.

The World Health Organization predicts that by the year 2030, road traffic injuries will be one of the top five leading causes of death. Many of these deaths and injuries can be prevented by driving cars properly equipped with state-of-the-art safety and driver assistance systems. Some examples are auto-brake and auto-collision avoidance which are becoming more and more popular on the market today. A recent study by a Swedish insurance company has shown that on roadswith speeds up to 50 km/h an auto-brake system can reduce personal injuries by up to 64 percent. In fact in an estimated 40 percent of crashes, the auto-brake reduced the effects to the degree that no personal injury was sustained. It is imperative that these so called Advanced Driver Assistance Systems, to be really effective, have good situational awareness. It is important that they have adequate information of the vehicle’s immediate surroundings. Where are other cars, pedestrians or motorcycles relative to our own vehicle? How fast are they driving and in which lane? How is our own vehicle driving? Are there objects in the way of our own vehicle’s intended path? These and many more questions can be answered by a properly designed system for situational awareness. In this thesis we design and evaluate, both quantitatively and qualitatively, sensor fusion algorithms for multi-target tracking. We use a combination of camera and radar information to perform fusion and find relevant objects in a cluttered environment. The combination of these two sensors is very interesting because of their complementary attributes. The radar system has high range resolution but poor bearing resolution. The camera system on the other hand has a very high bearing resolution. This is very promising, with the potential to substantially increase the accuracy of the tracking system compared to just using one of the two. We have also designed algorithms for path prediction and a first threat awareness logic which are both qualitively evaluated.

The problem considered in this thesis is the development of an autonomous prototype robot capable of gathering sensory information from its environment allowing it to provide feedback on the condition of specific targets to aid in maintenance of hydro equipment. The context for the solution to this problem is based on the power grid environment operated by the local hydro utility. The intent is to monitor power line structures by travelling along skywire located at the top of towers, providing a view of everything beneath it including, for example, insulators, conductors, and towers. The contribution of this thesis is a novel robot design with the potential to prevent hazardous situations and the use of rough coverage feedback modified reinforcement learning algorithms to establish behaviours. / October 2007

This thesis presents a rough set approach to reinforcement learning. This is made possible by considering behaviour patterns of learning agents in the context of approximation spaces. Rough set theory introduced by Zdzisław Pawlak in the early 1980s provides a ground for deriving pattern-based rewards within approximation spaces. Learning can be considered episodic. The framework provided by an approximation space makes it possible to derive pattern-based reference rewards at the end of each episode. Reference rewards provide a standard for reinforcement comparison as well as the actor-critic method of reinforcement learning. In addition, approximation spaces provide a basis for deriving episodic weights that provide a basis for a new form of off-policy Monte Carlo learning control method. A number of conventional and pattern-based reinforcement learning methods are investigated in this thesis. In addition, this thesis introduces two learning environments used to compare the algorithms. The first is a Monocular Vision System used to track a moving target. The second is an artificial ecosystem testbed that makes it possible to study swarm behaviour by collections of biologically-inspired bots. The simulated ecosystem has an ethological basis inspired by the work of Niko Tinbergen, who introduced in the 1960s methods of observing and explaining the behaviour of biological organisms that carry over into the study of the behaviour of interacting robotic devices that cooperate to survive and to carry out highly specialized tasks. Agent behaviour during each episode is recorded in a decision table called an ethogram, which records features such as states, proximate causes, responses (actions), action preferences, rewards and decisions (actions chosen and actions rejected). At all times an agent follows a policy that maps perceived states of the environment to actions. The goal of the learning algorithms is to find an optimal policy in a non-stationary environment. The results of the learning experiments with seven forms of reinforcement learning are given. The contribution of this thesis is a comprehensive introduction to a pattern-based evaluation of behaviour during reinforcement learning using approximation spaces. / May 2006

In order to understand the migration patterns of migrating birds, it is necessary to understand whenand where to they migrate. Many of these birds are very small and thus cannot carry heavy sensors;hence it is necessary to be able to perform positioning using a very small sensor. One way to do this isto use a light-intensity sensor. Since the sunrise and sunset times are known given time and position onthe earth, it is possible to determine the global position using light intensity. Light intensity increasesas the sun rises. Data sets from several calibration sensors, mainly from different locations in Sweden, have been examinedin different ways in order to get an understanding of the measurements and what affects them. Inorder to perform positioning, it is necessary to know the solar elevation angle, which can be computedif the time and position are known, as is the case for the calibration sensors. This has been utilized toidentify a mapping from measured light intensity to solar elevation angle, which is used to computepseudo-measurements for target tracking, described below. In this thesis, positioning is performed using methods from the field of target tracking. This is doneboth causally (filtering) and non-causally (smoothing). There are certain problems that arise; firstly,the measured light intensity can be attenuated due to weather conditions such as cloudiness, which ismodelled as a time-varying offset. Secondly, the sensor can be shadowed causing outliers in the data.Furthermore, birds are not always in a migratory state, they oftentimes stay in one place. The lattertwo phenomena are modelled using an Interacting Multiple Model (IMM) where they are representedas discrete states, corresponding to different models.

På överhettartuber i värmepannor bildas beläggningar på grund av sot och partiklar från förbränningen. Beläggningarna isolerar överhettartuberna vilket försämrar värmepannans verkningsgrad. Sotning av tuberna sker under drift och ett system som automatiskt kan beräkna hur mycket beläggningar som finns på tuberna skulle kunna göra sotningen mer behovsstyrd.Resonansfrekvenser hos överhettartuberna påverkas av masspåslaget som blir när beläggningar bildas. En förändring i frekvens kan då översättas till en förändring i massa. Vibrationsmätningar har gjorts med töjningsgivare som är monterade på överhettartuber i en av Ryaverkets pannor och i SAKAB:s panna. I detta examensarbete har syftet varit att analysera olika metoder för att skatta resonansfrekvenser i genomförda mätningar. Algoritmerna MUSIC, ESPRIT och AR samt en heuristisk statistisk metod har testats på genererade signaler. MUSIC och ESPRIT har givit bäst skattningar och har därefter använts för att analysera mätningarna. Periodvis följer vissa skattningar av mätningar från Ryaverket trender mellan sotningarna vilket indikerar att det skulle kunna vara resonansfrekvenser. Annars innehåller skattningarna för mycket variationer. Skattningarna av mätningar från SAKAB:s panna är gjorda under en kortare tidsperiod men visar en tydligare trend som troliggör att det är resonansfrekvenser som detekterats.För att automatiskt hitta och följa resonansfrekvenser i skattningarna, även om dessa innehåller stora variationer, har en målföljningsalgoritm implementerats. Algoritmen hittar skattningar som följer en förväntad trend mellan sotningarna. Tester visar att algoritmen hittar troliga resonansfrekvenser i skattningarna men att det är svårt att kunna dra några slutsatser om skattningarna varierar för mycket. Bättre signaler skulle kunna minska variationerna hos skattningarna. Ett förslag presenteras hur MUSIC eller ESPRIT tillsammans med en målföljningsalgoritm skulle kunna användas för att beräkna masspåslag på överhettartuberna. / Combustion in thermal power plants emits particles which create coatings on the super heater tubes. The coatings insulate the tubes and impairs the efficiency of the heat transfer. Cleaning the tubes occurs while the power plant is running and a system that automatically calculates the amount of coatings could make the cleaning more needs-based. The resonance frequencies of the super heater tubes are affected by the added mass of the coatings. A change in frequency corresponds to a change in mass. Vibration measurements have been made with strain gauges on the super heater tubes in Ryaverket's power plant and in one of SAKAB's power plants. The purpose of this thesis work has been to analyse different methods to estimate resonance frequencies in the vibration measurements. ESPRIT, MUSIC, AR and a heuristic statistical method have been tested on generated signals. MUSIC and ESPRIT have given the best estimations and have thus been used to analyse the measurements. Periodically some estimations of measurements from Ryaverket are following trends which indicates that they could be resonance frequencies. The rest of the estimations contain too large variations. The estimations made of the measurements from SAKAB have been made during a shorter time period but shows clearer trends which make them probable resonance frequencies.To automatically trace resonance frequencies in the estimations, even though they contains large variations, a target tracking algorithm has been implemented. The algorithm finds estimations that follows expected trends between the cleaning periods. Tests shows that the target tracking algorithm finds probable resonance frequencies in the estimations but that it is hard to reach a conclusion if they contain large variations. Better measurements could give estimations with smaller variations.An idea is presented where MUSIC or ESPRIT together with a target tracking algorithm could be used to calculate the amount of coatings on the super heater tubes.

This thesis discusses model based tracking of radar echoes in tanks with liquid content. The errors in the measurements are not always random in these cases, and the interference that occurs when nearby echoes moves in relation to each other makes tracking difficult. The tracking problem can be broken down to four parts; track initiation, track destruction, association of the measured echoes to the tracks and to update the tracks with the associated echoes. The radar measurement is primarily made on the surface. However, additional echoes from the tank floor or other structures and double bounces are normally present. There are often linear relations between the positions of the measured echoes. Kalman filters are typically used to estimate the positions of the echoes. Best performance is achieved if a model of the entire tank is used; however that model must be tailored to match the exact conditions in the tank. It is also possible to consider the echoes to be independent and track their positions using separate models. Some of the possiblemethods to automatically build a model for a specific tank have been explored.

Denna rapport behandlar en fullständig metod för att följa, associera och klassificera ekospår. Vid radarnivåmätning ska ekon som härstammar ifrån samma fysikaliska objekt följas över tiden och klassificeras. Dagens radarnivågivare löser problemet genom konfiguration. Tankens geometri samt vilka störande objekt som finns anges vid installation. Den framtagna metoden har delats upp i tre delmoment. Först binds individuella inmätta ekon samman till linjesegment. Dessa segment associeras samman till ekospår varefter dessa klassas fysikaliskt. Metoden kräver betydligt mindre konfiguration än den lösning som används för tillfället. / In this report a method for target tracking, association and classification of echo tracks in radar level gauges is discussed. In tank gauging echoes from the same physical object are called an echo track. The echoes must be tracked over time and classified. Currently radar level gauge units requires configuration to achieve classifications of echo tracks. At installation tank geometry and disturbing echoes has to be registered. The method has been divided into three parts. Firstly individual measured echoes are replaced by line segments. Association algorithms then create echo tracks from these segments. The third step is to classify these tracks physically. This method eliminates registering of disturbing echoes.

The problem of dynamic autonomous agent placement for tracking moving targets arises in many real-life applications, such as rescue operations, security, surveillance, and reconnaissance. The objective of this thesis is to develop a distributed hierarchical approach to address this problem. After the approach is developed, it is tested on a number of urban surveillance scenarios. The proposed approach views the placement problem as a multi-tiered architecture entailing modules for low-level sensor data preprocessing and fusion, decentralized decision support, knowledge building, and centralized decision support. This thesis focuses upon the modules of decentralized decision support and knowledge building. The decentralized decision support module requires a great deal of coordination among agents to achieve the mission objectives. The module entails two classes of distributed algorithms: non-model-based algorithms and model-based algorithms. The first class is used as a place holder while a model is built to describe agents knowledge about target behaviors. After the model is built and evaluated, agents switch to the model-based algorithms. To apply the approach to urban environments, urban terrain zones are classified, and the problem is mathematically formulated for two different types of urban terrain, namely low-rise, widely spaced and high-rise, closely spaced zones. An instance of each class of algorithms is developed for each of the two types of urban terrain. The algorithms are designed to run in a distributed fashion to address scalability and fault tolerance issues. The class of model-based algorithms includes a distributed model-based algorithm for dealing with evasive targets. The algorithm is designed to improve its performance over time as it learns from past experience how to deal with evasive targets. Apart from the algorithms, a model estimation module is developed to build motion models online from sensor observations. The approach is evaluated through a set of simulation experiments inspired from real-life scenarios. Experimental results reveal the superiority of the developed algorithms over existing ones and the applicability of the online model-building method. Therefore, it is concluded that the overall distributed approach is capable of handling agent placement or surveillance applications in urban environments among other applications.