Two crucial functionalities of a fully autonomous robotic agent are localization and navigation. The problem of enabling an agent to localize itself in an unknown environment is an extensive and widely studied topic. One of the main areas of this topic focuses on Simultaneous Localization and Mapping (SLAM). Many advancements in this field have been made over the years resulting in robust and accurate localization systems. Navigation progress has also improved substantially throughout the years resulting in efficient path planning algorithms and effective exploration strategies. Although an abundance of research exists on these two topics, less so exists about the combination of the two and their effect on each other. Therefore, the aim of this thesis was to integrate two state-of-the-art components from each respective area of research into a functioning system. This was done with the aim of studying the interconnectivity between these components while also documenting the integration process and identifying important considerations for similar future endeavours. Evaluations of the system showed that it performed with surprisingly good accuracy although it was severely lacking in robustness. Integration efforts showed good promise; however, it is clear that the two fields are heavily linked and need to be considered in a mutual context when it comes to a complete integrated system. / Förmågor som lokalisering och navigering är inom robotik förutsättande för att kunna möjliggöra en fullt autonom agent. Att för en agent kunna lokalisera sig i en okänd miljö är ett omfattande och brett studerat ämne, och ett huvudfokus inom ämnet är Simultaneous Localization and Mapping (SLAM) som avser lokalisering som sker parallellt med en aktiv kartläggning av omgivningen. Stora framsteg har gjorts inom detta område genom åren, vilket har resulterat i robusta och exakta system för robotlokalisering. Motsvarande framsteg inom robotnavigering har dessutom möjliggjort effektiva algoritmer och strategier för path planning och autonom utforskning. Trots den stora mängd forskning som existerar inom ämnena lokalisering och navigation var för sig, är samspelet mellan de två områdena samt möjligheten att sammankoppla de två aspekterna mindre studerat. I syfte att undersöka detta var målet med detta examensarbete således att integrera två toppmoderna system från de respektive områdena till ett sammankopplat system. Utöver att förmågorna och prestandan hos det integrerade systemet kunde studeras, genomfördes studien med avsikten att möjliggöra dokumentering av integrationsprocessen samt att viktiga insikter kring integrationen kunde identifieras i syfte att främja framtida studier inom samspelet mellan områdena lokalisering och navigation. Utvärderingar av det integrerade systemet påvisade en högre nivå av noggrannhet än förväntat, men fann en markant avsaknad av robusthet. Resultaten från integrationsarbetet anses lovande, och belyser framförallt att finns ett starkt samband mellan de två områdena samt att de bör beaktas i ett gemensamt kontext när de avses användas i ett komplett integrerat system.

Visual Simultaneous Localization And Mapping (V-SLAM) and object detection algorithms are two critical prerequisites for modern XR applications. V-SLAM allows XR devices to geometrically map the environment and localize itself within the environment, simultaneously. Furthermore, object detectors based on Deep Neural Network (DNN) can be used to semantically understand what those features in the environment represent. However, both of these algorithms are computationally expensive, which makes it challenging for them to achieve good real-time performance on device. In this thesis, we first present TensoRT Quantized YOLOv4 (TRTQYOLOv4), a faster implementation of YOLOv4 architecture [1] using FP16 reduced precision and INT8 quantization powered by NVIDIA TensorRT [2] framework. Second, we propose the V-SLAM Hurdler: A Faster VSLAM System using Online Dynamic-and-Hardness-aware Approximation. The proposed system integrates the base RGB-D V-SLAM ORB-SLAM3 [3] with the INT8 TRTQ-YOLOv4 object detector, a novel Entropy-based Degreeof- Difficulty Estimator, an Online Hardness-aware Approximation Controller and a Dynamic Object Eraser, applying online dynamic-and-hardness aware approximation to the base V-SLAM system during runtime while increasing its robustness in dynamic scenes. We first evaluate the proposed object detector on public object detection dataset. The proposed FP16 precision TRTQ-YOLOv4 achieves 2×faster than the full-precision model without loss of accuracy, while the INT8 quantized TRTQ-YOLOv4 is almost 3×faster than the full-precision one with only 0.024 loss in mAP@50:5:95. Second, we evaluate our proposed V-SLAM system on public RGB-D SLAM dataset. In static scenes, the proposed system speeds up the base VSLAM system by +21.2% on average with only −0.7% loss of accuracy. In dynamic scenes, the proposed system not only accelerate the base system by +23.5% but also improves the accuracy by +89.3%, making it as robust as in the static scenes. Lastly, the comparison against the state-of-the-art SLAMs designed dynamic environments shows that our system outperforms most of the compared methods in highly dynamic scenes. / Visual SLAM (V-SLAM) och objektdetekteringsalgoritmer är två kritiska förutsättningar för moderna XR-applikationer. V-SLAM tillåter XR-enheter att geometriskt kartlägga miljön och lokalisera sig i miljön samtidigt. Dessutom kan DNN-baserade objektdetektorer användas för att semantiskt förstå vad dessa egenskaper i miljön representerar. Men båda dessa algoritmer är beräkningsmässigt dyra, vilket gör det utmanande för dem att uppnå bra realtidsprestanda på enheten. I det här examensarbetet presenterar vi först TRTQ-YOLOv4, en snabbare implementering av YOLOv4 arkitektur [1] med FP16 reducerad precision och INT8 kvantisering som drivs av NVIDIA TensorRT [2] ramverk. För det andra föreslår vi V-SLAM-häckaren: ett snabbare V-SLAM-system som använder online-dynamisk och hårdhetsmedveten approximation. Det föreslagna systemet integrerar basen RGB-D V-SLAM ORB-SLAM3 [3] med INT8 TRTQYOLOv4 objektdetektorn, en ny Entropi-baserad svårighetsgradsuppskattare, en online hårdhetsmedveten approximationskontroller och en Dynamic Object Eraser, applicerar online-dynamik- och hårdhetsmedveten approximation till bas-V-SLAM-systemet under körning samtidigt som det ökar dess robusthet i dynamiska scener. Vi utvärderar först den föreslagna objektdetektorn på datauppsättning för offentlig objektdetektering. Den föreslagna FP16 precision TRTQ-YOLOv4 uppnår 2× snabbare än fullprecisionsmodellen utan förlust av noggrannhet, medan den INT8 kvantiserade TRTQ-YOLOv4 är nästan 3× snabbare än fullprecisionsmodellen med endast 0.024 förlust i mAP@50:5:95. För det andra utvärderar vi vårt föreslagna V-SLAM-system på offentlig RGB-D SLAM-datauppsättning. I statiska scener snabbar det föreslagna systemet upp V-SLAM-bassystemet med +21.2% i genomsnitt med endast −0.7% förlust av noggrannhet. I dynamiska scener accelererar det föreslagna systemet inte bara bassystemet med +23.5% utan förbättrar också noggrannheten med +89.3%, vilket gör det lika robust som i de statiska scenerna. Slutligen visar jämförelsen med de senaste SLAM-designade dynamiska miljöerna att vårt system överträffar de flesta av de jämförda metoderna i mycket dynamiska scener.

This study’s aim was to investigate the mapping part of Simultaneous Localisation And Mapping (SLAM) in indoor environments containing error sources relevant to two types of sensors. The sensors used were an Intel Realsense depth camera and an RPlidar Light Detection AndRanging (LiDAR). Both cameras and LiDARs are frequently used as exteroceptive sensors in SLAM. Cameras typically struggle with strong light in the environment, and LiDARs struggle with reflective surfaces. Therefore, this study investigated the possibility of using a neural network to detect an error in either sensors’ data caused by mentioned error sources. The network identified which sensor produced erroneous data. The sensor fusion algorithm momentarily excluded said sensor’s data, consequently, improving the mapping quality when possible. The quantitative results showed no significant difference in the measured mean squared error and structural similarity between the final maps generated with and without the network, when compared to the ground truth. However, the qualitative analysis showed some advantages with using the network. Many of the camera’s errors were filtered out with the neural network, and led to a more accurate continuous mapping than without the network implemented. The conclusion was that a neural network can to a limited extent recognise the sensors’ data errors, but only the camera data benefited from the proposed solution. The study also produced important findings from the implementation which are presented. Future work recommendations include neural network optimisation, sensor selection, and sensor fusion implementation. / Denna studie undersökte kartläggningen i Simultaneous Localisation And Mapping (SLAM) problem, i kontexten av två sensorers felkällor. Sensorerna som användes var en Intel Realsense djupseende kamera samt en LiDAR fran RPlidar. Både kameror och LiDARs är vanliga sensorer i SLAM system, och båda har olika typer av felkällor. Kameror är typiskt känsliga för mycket starkt ljus, medan LiDARs har svårt med reflekterande ytor. Med detta som bakgrund har denna studie undersökt möjligheten att implementera ett neuralt nätverk för att detektera när varje sensor är utsatt för en felkälla (och därmed ger fel data). Nätverkets klassificering används sedan för att i varje tidssteg exkludera den sensors data som det är fel på för att förbättra kartläggningen. De qvantitativa resultaten visade ingen signifikant skillnad mellan kartorna genererade med nätverket och de utan nätverket. Dock visade den kvalitativa analysen att det finns vissa fördelar med att använda det neutrala nätverket. Manga av kamerans fel blev korrigerade när nätverket var implementerat, vilket ledde till mer korrekta kartor under kontinuerlig körning. Slutsatsen blev att ett nätverk kan bli tränat för att identifiera fel i datan, men att kameran drar mest nytta av det. Studien producerade även sekundara resultat som också redovisas. Slutligen rekommenderas optimering av nätverket, val av sensorer, samt uppdaterad algoritm för sensor fusionen som möjliga områden till fortsatt forskning inom området.

Detta är en fallstudie där två olika Laserskannrar jämförs. Dessa skannrar skiljer sig då de använder två skilda metoder för punktmolnsregistrering. Dessa två metoder är: SLAM (Simultaneous Localization and Mapping) -baserad punktmolnsregistrering och punktmolnsregistrering med den indirekta två-stegs-metoden. Målet med projektet var således att granska den statiska SLAM-baserade skannern Leica RTC 360 3D och jämföra den med den mer traditionsenliga terrestra laserskannern Leica Scanstation P40. Denna undersökning är relevant eftersom det finns en tydlig skillnad mellan dessa laserskannrar både ur en planerings- och en effektivitetssynpunkt. Den statiska SLAM skannern RTC 360 har möjligheten att vara väldigt tidseffektiv då den använder sig av SLAM algoritmen VIS (Visual Inertial System) för punkmolnsregistrering i fält, samt att skannern använder en IMU (Inertial Measurement Unit) som möjliggör skanningar med hög kvalitet utan att instrumentet är ordentligt horisonterat. Vilket är en förutsättning för att kunna genomföra en skanning med P40. Punktmolnen från dessa två laserskannrar jämförs med varandra och granskas visuellt för att analysera vilka skillnader som finns mellan dem. Samma punktäthet används och skanningarna genomförs från samma positioner med båda instrumenten. Utöver detta görs en statistisk undersökning med hjälp av kontrollpunkter och RMS-värden beräknas med hjälp av dessa. För att tillåta denna statistiska analys, georefererades båda punktmolnen med indirekt georeferering till referenssystemet SWEREF 99 16 30 i plan samt RH 2000 i höjd. Just denna process var något enklare att genomföra med P40 än med RTC 360. Då P40 data var något bättre anpassat för bearbetning i Leica Cyclone. Vilket var de program som främst användes för bearbetning av data. RTC 360 använder två pulser för att skanna in varje punkt, vilket resulterade i att den skannade in något fler punkter. Framförallt märktes denna funktion på områden som traditionellt är svåra att skanna med en laserskanner, främst blanka, svarta områden. Det visade sig av kontrollpunkterna att det fanns en liten variation mellan punktmolnenskvalitet där lasersdata från P40 höll 1–2 millimeters kvalitet. Ur detta resultat kunde slutsatsen dras att Leica Scanstation P40 är något bättre för geodetiska ändamål med mycket höga krav på georefereringen. I det flesta andra sammanhang är RTC 360 att rekommendera. Leicas RTC 360 är ett bra exempel på hur SLAM-algoritmen kan användas för att förenkla många laserskanningsprojekt. Även i de projekten med höga krav på detaljrikedom. / This is a case study which aims to compare two different laser scanners. The main difference between these two scanners is that they use different solutions for registration of point clouds. These two solutions are SLAM (Simultaneous Localization and Mapping) as well as the indirect two-step approach. The thesis aims at comparing the static SLAM-based scanner Leica RTC 360 with the more traditional scanner Leica Scanstation P40. This is a relevant study due to the big differences between these two scanners, in the aspect of both how much preparation that is need and how effective both scanners are. The static SLAM scanner RTC 360 has the possibility to be very time efficient due to use of the SLAM-algorithm called VIS (Visual Inertial System) that are used for alignment of different point clouds as early as in the fieldwork. The RTC 360 also uses an IMU (Inertial Measurement Unit) that allows the laser scanner to do complete and detailed scans without the need to be perfectly levelled. The P40 on the other hand do need to be precisely levelled to be able to complete a scan. The point clouds from these two laser scanners are compared with each other by reviewing the visual features of the two clouds and finding differences between the point clouds. The same point density was used in both clouds and the scans took place from the same positions with both scanners. A statistical comparison was also made. This statistical analysis was made with use of control points and RMS values that were established with the help of these. This statistical analysis was made possible by the fact that both point clouds were georeferenced to the reference system SWEREF 99 16 30 as well the system RH 2000 for height. This process of georeferencing both clouds was easier to perform with the P40 than the RTC 360. Because the laser data from P40 were slightly better suited for the program Leica Cyclone, which were the program that was used for most of the data processing. RTC 360 uses two individual laser-pulses for each scanned point. This resulted in that the RTC 360 scanned some more points compared to the P40. This difference was extra noticeable on surfaces that usually are difficult for laser scanners to scan, such as plain, black surfaces. The control points showed that quality of both point clouds was very similar to each other. The P40 showed slightly higher accuracy, about 1-2millimeter, relative to the RTC 360 scanner. This resulted in the conclusion that P40 were slightly better for geodic purposes with very high demands on the georeferencing. In most of the other cases RTC 360 is the recommended scanner. Leica RTC 360 is a good example of how the SLAM-algorithm can be used to make many laser scanning projects easier and more efficient.

[pt] Essa dissertação tem por objetivo entender os sentidos e os efeitos políticos da participação feminina no slam poetry brasileiro, bem como expor e refletir sobre como essa forma de manifestação artística se relaciona com os espaços da cidade. O objetivo geral da reflexão situa-se no esforço de produzir uma leitura das apropriações do espaço urbano a partir dos eventos promovidos pelo Coletivo Slam das Minas (RJ), tendo por intenção pensar as articulações possíveis entre os debates levantados pelo campo temático da geografia e gênero, com formas criativas de resistência na cidade. Busca-se com isso, alcançar um entendimento aproximado sobre como mulheres e pessoas LGBTQIA+ estão atuando politicamente, por meio da poesia falada (spoken word) e da união de seus corpos em espaços públicos da cidade do Rio de Janeiro através dos eventos de Slam. / [en] This dissertation aims to understand the meanings and political effects of female participation in Brazilian slam poetry, as well as to expose and reflect on how this form of artistic manifestation relates to the spaces of the city. The general objective of the reflection lies in the effort to produce a reading of the appropriations of urban space from the events promoted by Coletivo Slam das Minas (RJ), with the intention of thinking about the possible articulations between the debates raised by the thematic field of geography and gender, with creative forms of resistance in the city. The aim is to achieve an approximate understanding of how women and LGBTQIA+ people are acting politically, through spoken word and the union of their bodies in public spaces in the city of Rio de Janeiro through Slam events

This thesis covers the topic of image matching in a visual SLAM or SfM context. These problems are generally based on a vector representation of the keypoints of one image, called a descriptor, which we seek to map to the keypoints of another, using a similarity measure to compare the descriptors. However, it remains difficult to perform this matching successfully, especially for challenging scenes where illumination changes, occlusions, motion, textureless and similar features are present, leading to mis-matched points. In this thesis, we develop a self-supervised contrastive deep learning framework for computing robust descriptors, particularly for these challenging situations.We use the TartanAir dataset built explicitly for this task, and in which these difficult scene cases are present. Our results show that descriptor learning works, improves scores, and that our method is competitive with traditional methods such as ORB. In particular, the invariance built implicitly by training pairs of positive examples through the construction of a trajectory from a sequence of images, as well as the controlled introduction of ambiguous negative examples during training, have a real observable effect on the scores obtained. / Le présent mémoire traite du sujet de mise en correspondance entre deux images dans un contexte de SLAM visuel ou de SfM. Ces problèmes reposent généralement sur une représentation vectorielle de points saillants d’une image, appelée descripteur, et qu’on cherche à mettre en correspondance avec les points saillants d’une autre, en utilisant une mesure de similarité pour comparer les descripteurs. Cependant, il reste difficile de réaliser cette mise en correspondance avec succès, en particulier pour les scènes difficiles où des changements d’illumination, des occultations, des mouvements, des éléments sans texture, et des éléments similaires sont présents, conduisant à des mises en correspondance incorrectes. Nous développons dans ce mémoire une méthode d’apprentissage profond contrastif auto-supervisé pour calculer des descripteurs robustes, particulièrement à ces situations difficiles. Nous utilisons le jeu de données TartanAir construit explicitement pour cette tâche, et dans lequel ces cas de scènes difficiles sont présents. Nos résultats montrent que l’apprentissage de descripteurs fonctionne, améliore les scores, et que notre méthode est compétitive avec les méthodes traditionnelles telles que ORB. En particulier, l’invariance bâtie implicitement en formant des paires d’exemples positifs grâce à la construction d’une trajectoire depuis une séquence d’images, ainsi que l’introduction contrôlée d’exemples négatifs ambigus pendant l’entraînement a un réel effet observable sur les scores obtenus.

Thesis (MSc)--Stellenbosch University, 2012. / ENGLISH ABSTRACT: In this thesis we consider the problem of building a dense and consistent map of a mobile robot’s environment that is updated as the robot moves. Such maps are vital for safe and collision-free navigation. Measurements obtained from a range sensor mounted on the robot provide information on the structure of the environment, but are typically corrupted by noise. These measurements are also relative to the robot’s unknown pose (location and orientation) and, in order to combine them into a world-centric map, pose estimation is necessary at every time step. A SLAM system can be used for this task. However, since landmark measurements and robot motion are inherently noisy, the pose estimates are typically characterized by uncertainty. When building a map it is essential to deal with the uncertainties in range measurements and pose estimates in a principled manner to avoid overconfidence in the map. A literature review of robotic mapping algorithms reveals that the occupancy grid mapping algorithm is well suited for our goal. This algorithm divides the area to be mapped into a regular lattice of cells (squares for 2D maps or cubes for 3D maps) and maintains an occupancy probability for each cell. Although an inverse sensor model is often employed to incorporate measurement uncertainty into such a map, many authors merely state or depict their sensor models. We derive our model analytically and discuss ways to tailor it for sensor-specific uncertainty. One of the shortcomings of the original occupancy grid algorithm is its inability to convey uncertainty in the robot’s pose to the map. We address this problem by altering the occupancy grid update equation to include weighted samples from the pose uncertainty distribution (provided by the SLAM system). The occupancy grid algorithm has been criticized for its high memory requirements. Techniques have been proposed to represent the map as a region tree, allowing cells to have different sizes depending on the information received for them. Such an approach necessitates a set of rules for determining when a cell should be split (for higher resolution in a local region) and when groups of cells should be merged (for lower resolution). We identify some inconsistencies that can arise from existing rules, and adapt those rules so that such errors are avoided. We test our proposed adaptive occupancy grid algorithm, that incorporates both measurement and pose uncertainty, on simulated and real-world data. The results indicate that these uncertainties are included effectively, to provide a more informative map, without a loss in accuracy. Furthermore, our adaptive maps need far fewer cells than their regular counterparts, and our new set of rules for deciding when to split or merge cells significantly improves the ability of the adaptive grid map to mimic its regular counterpart. / AFRIKAANSE OPSOMMING: In hierdie tesis beskou ons die probleem om ’n digte en konsekwente kaart van ’n mobiele robot se omgewing te bou, wat opgedateer word soos die robot beweeg. Sulke kaarte is van kardinale belang vir veilige, botsingvrye navigasie. Metings verkry vanaf ’n sensor wat op die robot gemonteer is, verskaf inligting rakende die struktuur van die omgewing, maar word tipies deur ruis vervorm. Hierdie metings is ook relatief tot die robot se onbekende postuur (posisie en oriëntasie) en, om hulle saam te voeg in ’n wêreldsentriese kaart, is postuurafskatting nodig op elke tydstap. ’n SLAM stelsel kan vir hierdie doeleinde gebruik word. Aangesien landmerkmetings en die beweging van die robot inherent ruiserig is, word die postuurskattings gekarakteriseer deur onsekerheid. Met die bou van ’n kaart moet hierdie onsekerhede in afstandmetings en postuurskattings op ’n beginselvaste manier hanteer word om te verhoed dat te veel vertroue in die kaart geplaas word. ’n Literatuurstudie van karteringsalgoritmes openbaar die besettingsroosteralgoritme as geskik vir ons doel. Die algoritme verdeel die gebied wat gekarteer moet word in ’n reëlmatige rooster van selle (vierkante vir 2D kaarte of kubusse vir 3D kaarte) en onderhou ’n besettingswaarskynlikheid vir elke sel. Alhoewel ’n inverse sensormodel tipies gebruik word om metingsonsekerheid in so ’n kaart te inkorporeer, noem of wys baie outeurs slegs hulle model. Ons herlei ons model analities en beskryf maniere om sensorspesifieke metingsonsekerheid daarby in te sluit. Een van die tekortkominge van die besettingsroosteralgoritme is sy onvermoë om onsekerheid in die postuur van die robot na die kaart oor te dra. Ons spreek hierdie probleem aan deur die opdateringsvergelyking van die oorspronklike besettingsroosteralgoritme aan te pas, om geweegde monsters van die postuuronsekerheidsverdeling (verskaf deur die SLAM stelsel) in te sluit. Die besettingsroosteralgoritme word soms gekritiseer vir sy hoë verbruik van geheue. Tegnieke is voorgestel om die kaart as ’n gebiedsboom voor te stel, wat selle toelaat om verskillende groottes te hê, afhangende van die inligting wat vir hulle verkry is. So ’n benadering noodsaak ’n stel reëls wat spesifiseer wanneer ’n sel verdeel (vir ’n hoër resolusie in ’n plaaslike gebied) en wanneer ’n groep selle saamgevoeg (vir ’n laer resolusie) word. Ons identifiseer teenstrydighede wat kan voorkom as die huidige reëls gevolg word, en pas hierdie reëls aan sodat sulke foute vermy word. Ons toets ons voorgestelde aanpasbare besettingsroosteralgoritme, wat beide metings- en postuuronsekerheid insluit, op gesimuleerde en werklike data. Die resultate dui daarop dat hierdie onsekerhede op ’n effektiewe wyse na die kaart oorgedra word sonder om akkuraatheid prys te gee. Wat meer is, ons aanpasbare kaarte benodig heelwat minder selle as hul reëlmatige eweknieë. Ons nuwe stel reëls om te besluit wanneer selle verdeel of saamgevoeg word, veroorsaak ook ’n merkwaardige verbetering in die vermoë van die aanpasbare roosterkaart om sy reëlmatige eweknie na te boots.

This thesis focuses on the theory and implementation of visual navigation techniques for Autonomous Air Vehicles in outdoor environments. The target of this study is to fuse and cooperatively develop an incremental map for multiple air vehicles under the application of Simultaneous Location and Mapping (SLAM). Without loss of generality, two unmanned air vehicles (UAVs) are investigated for the generation of ground maps from current and a priori data. Each individual UAV is equipped with inertial navigation systems and external sensitive elements which can provide the possible mixture of visible, thermal infrared (IR) image sensors, with a special emphasis on the stereo digital cameras. The corresponding stereopsis is able to provide the crucial three-dimensional (3-D) measurements. Therefore, the visual aerial navigation problems tacked here are interpreted as stereo vision based SLAM (vSLAM) for both single and multiple UAVs applications. To begin with, the investigation is devoted to the methodologies of feature extraction. Potential landmarks are selected from airborne camera images as distinctive points identified in the images are the prerequisite for the rest. Feasible feature extraction algorithms have large influence over feature matching/association in 3-D mapping. To this end, effective variants of scale-invariant feature transform (SIFT) algorithms are employed to conduct comprehensive experiments on feature extraction for both visible and infrared aerial images. As the UAV is quite often in an uncertain location within complex and cluttered environments, dense and blurred images are practically inevitable. Thus, it becomes a challenge to find feature correspondences, which involves feature matching between 1st and 2nd image in the same frame, and data association of mapped landmarks and camera measurements. A number of tests with different techniques are conducted by incorporating the idea of graph theory and graph matching. The novel approaches, which could be tagged as classification and hypergraph transformation (HGTM) based respectively, have been proposed to solve the data association in stereo vision based navigation. These strategies are then utilised and investigated for UAV application within SLAM so as to achieve robust matching/association in highly cluttered environments. The unknown nonlinearities in the system model, including noise would introduce undesirable INS drift and errors. Therefore, appropriate appraisals on the pros and cons of various potential data filtering algorithms to resolve this issue are undertaken in order to meet the specific requirements of the applications. These filters within visual SLAM were put under investigation for data filtering and fusion of both single and cooperative navigation. Hence updated information required for construction and maintenance of a globally consistent map can be provided by using a suitable algorithm with the compromise between computational accuracy and intensity imposed by the increasing map size. The research provides an overview of the feasible filters, such as extended Kalman Filter, extended Information Filter, unscented Kalman Filter and unscented H Infinity Filter. As visual intuition always plays an important role for humans to recognise objects, research on 3-D mapping in textures is conducted in order to fulfil the purpose of both statistical and visual analysis for aerial navigation. Various techniques are proposed to smooth texture and minimise mosaicing errors during the reconstruction of 3-D textured maps with vSLAM for UAVs. Finally, with covariance intersection (CI) techniques adopted on multiple sensors, various cooperative and data fusion strategies are introduced for the distributed and decentralised UAVs for Cooperative vSLAM (C-vSLAM). Together with the complex structure of high nonlinear system models that reside in cooperative platforms, the robustness and accuracy of the estimations in collaborative mapping and location are achieved through HGTM association and communication strategies. Data fusion among UAVs and estimation for visual navigation via SLAM were impressively verified and validated in conditions of both simulation and real data sets.

Robust long-term positioning for autonomous mobile robots is essential for many applications. In many environments this task is challenging, as errors accumulate in the robot’s position estimate over time. The robot must also build a map so that these errors can be corrected when mapped regions are re-visited; this is known as Simultaneous Localisation and Mapping, or SLAM. Successful SLAM schemes have been demonstrated which accurately map tracks of tens of kilometres, however these schemes rely on expensive sensors such as laser scanners and inertial measurement units. A more attractive, low-cost sensor is a digital camera, which captures images that can be used to recognise where the robot is, and to incrementally position the robot as it moves. SLAM using a single camera is challenging however, and many contemporary schemes suffer complete failure in dynamic or featureless environments, or during erratic camera motion. An additional problem, known as scale drift, is that cameras do not directly measure the scale of the environment, and errors in relative scale accumulate over time, introducing errors into the robot’s speed and position estimates. Key to a successful visual SLAM system is the ability to continue operation despite these difficulties, and to recover from positioning failure when it occurs. This thesis describes the development of such a scheme, which is known as BoWSLAM. BoWSLAM enables a robot to reliably navigate and map previously unknown environments, in real-time, using only a single camera. In order to position a camera in visually challenging environments, BoWSLAM combines contemporary visual SLAM techniques with four new components. Firstly, a new Bag-of-Words (BoW) scheme is developed, which allows a robot to recognise places it has visited previously, without any prior knowledge of its environment. This BoW scheme is also used to select the best set of frames to reconstruct positions from, and to find efficient wide-baseline correspondences between many pairs of frames. Secondly, BaySAC, a new outlier- robust relative pose estimation scheme based on the popular RANSAC framework, is developed. BaySAC allows the efficient computation of multiple position hypotheses for each frame. Thirdly, a graph-based representation of these position hypotheses is proposed, which enables the selection of only reliable position estimates in the presence of gross outliers. Fourthly, as the robot explores, objects in the world are recognised and measured. These measurements enable scale drift to be corrected. BoWSLAM is demonstrated mapping a 25 minute 2.5km trajectory through a challenging and dynamic outdoor environment in real-time, and without any other sensor input; considerably further than previous single camera SLAM schemes.

Vid tillverkning av massa och papper förorenas årligen 505 miljoner kubikmeter vatten som måste renas innan det släpps tillbaka till omgivningen. Vid reningen avskiljs först stora partiklar som sedan avvattnas och förbränns. Vattnet som blir kvar genomgår ytterligare en rening, varvid det bildas bioslam. Bioslammet innehåller mycket intracellulärt vatten, vilket gör det kostsamt och energikrävande att avvattna. Det är även sedan 2005 förbjudet att dumpa organiskt material, varför en mer ekonomiskt attraktiv behandling av slammet är anaerob nedbrytning. I denna nedbrytning omvandlas det organiska materialet till metan och koldioxid där metanet är den eftertraktade gasen. Bioslammet innehåller dock partiklar såsom träfiberrester och mikroorganismer med komplex struktur och är näringsfattigt. Någon form av sönderdelande förbehandling underlättar därför rötningsprocessen. I detta projekt undersöktes termisk förbehandling i kombination med kemisk förbehandling på bioslam från Stora Enso Skoghalls bruk på Hammarö. Själva rötningen skedde i två omgångar varav den första omgången med termisk förbehandling vid 70C och den andra vid 140C. Den kemiska förbehandlingen skedde med tillsats av lut (natriumhydroxid), kalk (kalciumhydroxid) och syra (fosforsyra) vid pH 9 och 11 för baserna och pH 2 och 4 för syran. Även neutrala prov (endast värmebehandling) och ett blankprov (ingen förbehandling) gjordes. Bioslammet ympades med kommunalt slam från Fiskartorpets reningsverk i Kristinehamn som har en mesofil bakteriekultur. Rötningen varade i 19 dagar per omgång i en temperatur på 35C och skedde satsvis i E-kolvar försedda med påsar för gasuppsamling. Totalt rötades 42 prov per omgång som utgjordes av sju mätpunkter á sex replikat för goda statistiska underlag. Resultaten gav en indikation för högst metanproduktion för proven behandlade med kalk vid 140C och för provet utan kemisk förbehandling vid 140C. Lägst produktion hade det kalkbehandlade provet vid pH 9 och 70C följt av blankprovet. Lutproven gav lägre metanproduktion vid 140C än vid 70C och fosforsyran hade så gott som oförändrad produktion mellan temperaturerna. Gemensamt för alla prover som behandlats vid 70C var att de fick en högre procentandel metan då de behandlats vid 140C. De resultat som erhållits är dock osäkra då det i vissa fall var stor spridning mellan provens biogasproduktion inom de enskilda förbehandlingsområdena. / In the pulp and paper process 505 million tons of water are polluted annually, which has to be purified before it is returned to the surrounding lakes. When the water is treated bigger particles are first separated to form sludge, then dewatered and finally incinerated. The excess water is further treated were a type of sludge  bio sludge  is formed. The bio sludge contains high concentration of intracellular water, why it is expensive and energy demanding to dewater. It is also forbidden to dump organic waste since 2005, why a more economically attractive treatment of the water is anaerobic digestion. In the digestion organic compounds is converted into methane and carbon dioxide where the methane is the desired gas. The bio sludge also contains fiber residues and microorganisms with complex structure and is nutrient-poor, which makes it hard to digest. Some kind of disintegrating pretreatment is needed and co-digestion with a more nutrient-rich sludge to facilitate the digestion process. In this project thermal pretreatment in combination with chemical pretreatment was examined on bio sludge from Stora Enso Skoghalls bruk at Hammarö. The anaerobe digestion was done by two rounds whereof the first round thermal pretreated at 70C and the second at 140C. The chemical pretreatment was done by additive of sodium hydroxide, calcium hydroxide and phosphoric acid at pH 9 and 11 for the bases and pH 2 and 4 for the acid. Also neutral samples (no chemical pretreatment) and a reference sample (no pretreatment) were done. The bio sludge were co-digested with municipal sludge from Fiskartorpets reningsverk in Kristinehamn which has a mesophilic bacterial culture. The anaerobic digestion lasted for 19 days per round at a temperature of 35C and were done batch wise in E-flasks provided with a small bag for gas collection. Totally 42 samples were made per round which consisted of seven measurement points and six replicates each for a good statistical basis. The results gave an indication of the highest methane production for the samples treated with calcium hydroxide at 140C and the neutral sample treated at 140C. The sample treated with calcium hydroxide at pH 9 and 70C gave the lowest production of methane followed by the reference sample. The samples treated with sodium hydroxide gave a lower methane production at 140C than at 70C while the acid treated samples had almost the same production at the two different temperatures. All the samples had in common a higher proportion of methane in the biogas when treated at 140C than at 70C. The results should be taken with caution since the distribution amongst the samples within the same pretreatment method sometimes is very high.