Browsing by Author "Brahmanage, Gayan Sampath"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
- ItemOpen AccessGeometric Feature Based Rao-Blackwellized Particle Filter SLAM Using RGB-D Depth Images(2018-01-18) Brahmanage, Gayan Sampath; Leung, Henry; Macnab, Chris; Ramirez-Serrano, AlejandroIn recent years, commercially available mobile robots that operate in indoor environments have become more popular in household, office environment and industrial settings. These robots are used to provide services autonomously such as cleaning and surveillance. Simultaneous Localization And Mapping (SLAM) is a framework that can assist robots to navigate autonomously. SLAM builds the map of the environment where the robot operates and simultaneously localizes the robot in that environment. To develop SLAM, many approaches have emerged using high precision and long-range sensors such as laser scanners. However, laser scanners are very expensive and provide no visual information. In this thesis, novel approaches to build maps using low cost RGB-D sensors in indoor environments are investigated. A goal of this thesis is to consider a major limitation on the hardware side of commercially available low-cost robots, namely processing power and available memory. A new approach is developed for RGB-D cameras that captures 2D maps by processing reduced amount of data acquired from RGB-D data frames. This approach uses distinct geometric features that are extracted from depth data in a 2D plane. Further, the performances of grid-based mapping is exploited by replacing laser scanners with RGB-D camera. Since RGB-D cameras provide both color and depth data, it is possible to fuse these two data frames to exploit the advantages over using laser scanners. The stored RGB images at each pose can be used to add additional capabilities to 2D SLAM such as object recognition. In this context, this work is further extended to construct a 3D map of the environment using estimated 2D poses and stored sequence of RGB-D data frames using RGB features. Methods are evaluated for accuracy and consistency using experimental data gathered from real environments. Also, the proposed approach is implemented to build 2D maps in real time on a low-cost robot.
- ItemOpen AccessRGB Predicted Depth Simultaneous Localization and Mapping (SLAM) for Outdoor Environment(2024-04-18) Brahmanage, Gayan Sampath; Leung, Henry; Wang, Yingxu; Hu, Yaoping; Bisheban, Mahdis; Gu, JasonThis thesis focuses on visual simultaneous localization and mapping (V-SLAM) for outdoor applications such as autonomous driving. While most V-SLAM methods have been tested on small-scale settings such as mobile robots, applying them in expansive outdoor spaces introduces additional complexities. The larger scale of the environment, dynamic obstacles, and depth-perception limitations of visual sensors pose challenges for V-SLAM methods. The first contribution introduces a dynamic V-SLAM approach. A novel front-end motion tracking approach is developed to recover multiple motions from image frames, considering key-points observed after map initialization as dynamic with time-varying locations. The proposed approach searches for key-point clusters based on their motion and classifies associated motions probabilistically. A bundle adjustment (BA) optimizes the local map, camera trajectory, and key-points motion within a unified V-SLAM system. BA maintains the geometric relationships between dynamic key-points and camera poses in the co-visibility graph, enhancing the overall robustness and accuracy of V-SLAM in populated environments. The second contribution of this thesis centers around a deep-learning-based depth prediction approach, which proves effective for estimating metric scale maps using a monocular camera. An unsupervised depth prediction approach is proposed using a novel convolution vision transformer (CViT) model architecture to infer depth from monocular images. The proposed encoder features a dual CViT block (DCViT); one block generates self-attention solely based on the spatial context of input feature vectors, and the other learns to generate attention based on the scene’s geometry. Contrastive learning of visual representations is applied to DCViT, where the model takes depth predictions from the same model through a feedback path as a supervisory signal to train the DCViT. Integration with residual blocks enables the learning of local and global receptive fields that produce predicted disparity maps at a higher level of detail and accuracy. Experimental results demonstrate significant improvements over state-of-the-art methods across multiple depth datasets. The third contribution of this thesis involves a comprehensive investigation into the utilization of predicted depth within monocular SLAM. This exploration aims to enhance the accuracy of map estimation in metric scale. Most existing approaches struggle with the non-Gaussian distribution inherent in heavy-tail noise produced by depth prediction models. The proposed monocular SLAM approach utilizes t-distribution for ego-motion, with parameter estimation achieved through maximum likelihood (ML) estimation using the expectation maximization (EM) algorithm. Experiments on real data show that the proposed t-distribution renders the monocular SLAM algorithm inherently robust to outliers and heavy-tail noise produced by depth prediction models.