3D ranging sensor LiDAR, is an extensively used sensor in the autonomous vehicle industry, but LiDAR placement problem is not studied extensively. This thesis work proposes a framework in an open- source autonomous driving simulator (CARLA) that aims to solve LiDAR placement problem, based on the tasks that LiDAR is intended for in most of the autonomous vehicles. LiDAR placement problem is solved by improving point cloud density around the vehicle, and this is calculated by using LiDAR Occupancy Boards (LOB). Introducing LiDAR Occupancy as an objective function, the genetic algorithm is used to optimize this problem. This method can be extended for multiple LiDAR placement problem. Additionally, for multiple LiDAR placement problem, LiDAR scan registration algorithm (NDT) can also be used to find a better match for first or reference LiDAR. Multiple experiments are carried out in simulation with a different vehicle truck and car, different LiDAR sensors Velodyne 16 and 32 channel LiDAR, and, by varying Region Of Interest (ROI), for testing the scalability and technical robustness of the framework. Finally, this framework is validated by comparing the current and proposed LiDAR positions on the truck. / 3D- sensor LiDAR, är en sensor som används i stor utsträckning inom den autonoma fordonsindustrin, men LiDAR- placeringsproblemet studeras inte i stor utsträckning. Detta uppsatsarbete föreslår en ram i en öppen källkod för autonom körningssimulator (CARLA) som syftar till att lösa LiDAR- placeringsproblem, baserat på de uppgifter som LiDAR är avsedda för i de flesta av de autonoma fordonen. LiDAR- placeringsproblem löses genom att förbättra punktmolntätheten runt fordonet, och detta beräknas med LiDAR Occupancy Boards (LOB). Genom att introducera LiDAR Occupancy som en objektiv funktion används den genetiska algoritmen för att optimera detta problem. Denna metod kan utökas för flera LiDAR- placeringsproblem. Dessutom kan LiDAR- scanningsalgoritm (NDT) för flera LiDAR- placeringsproblem också användas för att hitta en bättre matchning för LiDAR för första eller referens. Flera experiment utförs i simulering med ett annat fordon lastbil och bil, olika LiDAR-sensorer Velodyne 16 och 32kanals LiDAR, och, genom att variera intresseområde (ROI), för att testa skalbarhet och teknisk robusthet i ramverket. Slutligen valideras detta ramverk genom att jämföra de nuvarande och föreslagna LiDAR- positionerna på lastbilen.

As the demand for 3D maps from LIDAR scanners increases, delivering high-quality maps becomes critical. One way to ensure the quality of such maps is through point cloud alignment classification, which aims to classify the alignment error between two registered point clouds. Specifically, we present the classifier FACT (Feature-Aware Classification Transformer), consisting of two main modules: feature extraction and classification. Descriptive features are extracted from the joint point cloud, which are then processed by a point transformer-based neural network to predict the alignment error class. In a ten-class point cloud alignment classification test, FACT achieved 92.4% accuracy, where the alignment error ranged from zero meters and radians to 0.9 meters and 0.09 radians. Remarkably, the classifier only made one misprediction beyond neighboring classes, exhibiting its ability to detect alignment errors as the classes have an inherent order. Furthermore, when benchmarked on two binary classification tasks, FACT showed significantly superior performance over the baseline and even obtained 100.0% accuracy for the easier of the two tasks. FACT not only detects potential errors in 3D maps but also estimates their magnitude, leading to more reliable 3D maps with quality estimations for each transformation.