Browsing by Subject "SLAM"

Now showing 1 - 8 of 8

A hierarchical multi-robot mapping architecture subject to communication constraints
(2020-12) Cappel, Henry Fielding; Sentis, Luis
Multi-robot systems are an efficient method to explore and map an unknown environment. The simulataneous localization and mapping (SLAM) algorithm is common for single robot systems, however multiple robots can share respective map data in order to merge a larger global map. This thesis contributes to the multi-robot mapping problem by considering cases in which robots have communication range limitations. The architecture coordinates a team of robots and the central server to explore an unknown environment by exploiting a hierarchical choice structure. The coordination algorithms ensure that the hierarchy of robots choose frontier points that provide maximum information gain, while maintaining viable communication amongst themselves and the central computer through an ad-hoc relay network. In addition, the robots employ a backup choice algorithm in cases when no valid frontier points remain by arranging the communication relay network as a fireline back to the source. This work contributes a scalable, efficient, and robust architecture towards hybrid multi-robot mapping systems that take into account communication range limitations. The architecture is tested in a simulation environment using various maps.
A component-oriented approach to simultaneous localization and mapping
(2008-05) Ristroph, Mickey; Browne, J.C.
The simultaneous localization and mapping (SLAM) problem is central to many mobile robots. The construction of a map of the local environment and the localization of the robot in that map must be accomplished in an incremental manner, even in the presence of significant error and uncertainty in sensor data. Further, the cyclic data dependency presents a challenge that does not lend itself to robust solutions. Solutions to the SLAM problem have been a major success in robotics research in the past 20 years. Existing solutions fully embrace the self-referential nature of the problem. There is a certain art to designing and tuning systems that achieve stability and convergence properties. Most robotics platforms implement a customized version of SLAM with modifications to improve performance, given certain assumptions and a priori knowledge of the environment. Borrowing techniques and tools from the parallel composition research community, we aimed to design a robust, extensible, and efficient framework for SLAM solutions using a component- based architecture. This begins with a domain analysis, characterizing the breadth of existing solutions and factoring the logical function of various components in a modular way. We then define the interfaces for these components, provide implementations, and connect them in a data ow or dependency graph. The final implementation presented supports, in theory, particle lter localization and can benefit from automated parallelization. However, it has not yet run successfully at the time of writing. The reasons for this include missing key features and non-working implementations of purported features in the tools used, PCOM2 and CODE. These limitations are described in detail and motivated, and should serve as justification for future work in parallel component composition research. Unfortunately, these are the only presentable results of the research done here at this time.
Direct monocular SLAM augmented with LIDAR range measurements
(2021-01-26) Marcus, Corey Leonard; Zanetti, Renato, 1978-
There are many situations where a spacecraft requires knowledge of the geometry and pose relative to the surrounding environment. SLAM is a class of algorithms capable of solving these problems simultaneously. In this thesis we present a Direct Monocular SLAM system which is augmented with flash LIDAR images. LIDAR measurements provide three tangible improvements to Direct Monocular SLAM. Improved SLAM system initialization through the use of LIDAR. Metric LIDAR measurements allow the true scale of localization and mapping estimates to become observable. And finally, an Extended Kalman filter is used to provide updates on map features using LIDAR images. Monte Carlo methods are used to demonstrate that incorporating LIDAR measurements into the system provides significant performance improvements over a system without LIDAR
High-Precision Globally-Referenced Position and Attitude via a Fusion of Visual SLAM, Carrier-Phase-Based GPS, and Inertial Measurements
(2014-05) Shepard, Daniel P.; Humphreys, Todd E.
A novel navigation system for obtaining highprecision globally-referenced position and attitude is presented and analyzed. The system is centered on a bundleadjustment-based visual simultaneous localization and mapping (SLAM) algorithm which incorporates carrierphase differential GPS (CDGPS) position measurements into the bundle adjustment in addition to measurements of point features identified in a subset of the camera images, referred to as keyframes. To track the motion of the camera in real-time, a navigation filter is employed which utilizes the point feature measurements from all non-keyframes, the point feature positions estimated by bundle adjustment, and inertial measurements. Simulations have shown that the system obtains centimeter-level or better absolute positioning accuracy and sub-degree-level absolute attitude accuracy in open outdoor areas. Moreover, the position and attitude solution only drifts slightly with the distance traveled when the system transitions to a GPS-denied environment (e.g., when the navigation system is carried indoors). A novel technique for initializing the globallyreferenced bundle adjustment algorithm is also presented which solves the problem of relating the coordinate systems for position estimates based on two disparate sensors while accounting for the distance between the sensors. Simulation results are presented for the globally-referenced bundle adjustment algorithm which demonstrate its performance in the challenging scenario of walking through a hallway where GPS signals are unavailable.
Multi-modal 3D Gaussian Splatting for SLAM
(2024) Sun, Lisong C.; Wang, Zhangyang
From AR/VR to autonomous mobile robotics, Simultaneous Localization and Mapping (SLAM) is essential for tracking and scene understanding. 3D Gaussian Splatting (3DGS) offers a map representation capable of photorealistic reconstruction and real-time rendering of scenes using multiple posed cameras. By combining these two techniques, this thesis aims to show that a 3D Gaussian map representation is capable of accurate SLAM when given unposed RGB-D images and inertial measurements. The proposed method, MM3DGS, addresses the limitations of prior neural radiance field representations by enabling faster rendering, scale awareness, and improved trajectory tracking. In addition, a new multi-modal SLAM dataset, UT-MM, is collected from a mobile robot and is publicly released. Experimental evaluation on several scenes from the dataset shows that with the proper sensor conf iguration, MM3DGS achieves 3× improvement in tracking and 5% improvement in photometric rendering quality compared to the current 3DGS SLAM state-of-the-art, while allowing real-time rendering of a high-resolution dense 3D map.
Observability Analysis of Opportunistic Navigation with Pseudorange Measurements
(2012-08-14) Kassas, Zak; Humphreys, Todd
Onboard control, tracking and navigation for autonomous systems
(2023-05) Ram S P, Arjun; Akella, Maruthi Ram, 1972-; Zanetti, Renato, 1978-; Clarke, John Paul; Wang, Junmin; Tanaka, Takashi
To operate effectively and safely, autonomous systems must be able to navigate complex environments, precisely control their attitude, accurately estimate their own states and make real-time decisions. A new adaptive controller is designed for attitude tracking control of rigid spacecraft with inertia uncertainties and full state feedback of attitude and angular rate. The controller preserves the proportional-derivative plus feedforward (PD+) structure but introduces time varying feedback gains, wherein the desired attitude state is represented by quaternions. Stable asymptotic tracking of the desired reference trajectories is guaranteed without any further restrictions upon the initial conditions, reference trajectories or any requirement for a priori availability of bounds upon the inertia matrix. The onboard estimation of the angular velocity of a spacecraft using Rate-integrating Gyroscopes (RIGs) is considered next. RIGs provide measurements of angular displacement which need to a low pass filter or observer to obtain angular velocity of the system. A continuous time observer is proposed which estimates angular velocity using the RIG measurements and achieves exponential convergence, while asymptotic convergence is guaranteed for the adaptive inertia observer. Unlike conventional certainty equivalence methods, a novel adaptation update law is proposed with additional control knobs changing with the attitude states. The final part of the dissertation deals with the Simultaneous Localization and Mapping (SLAM) problem. Estimating the location of a robot along with the position of the surrounding features on a map can be done onboard recursively with a simple Extended Kalman Filter (EKF-SLAM), but has higher chances of divergence and inconsistency. The robocentric SLAM method transforms the map of features to a local reference frame centered at the robot's position, leading to reduced inconsistency due to lower linearization errors in the update function. Improvements to the robocentric SLAM methods are suggested in the form of addition of second order terms to the linear propagation step and elimination of the composition step by transforming the feature maps before every update step. These modifications provide better filter consistency and prevent divergence in cases which were previously not possible.
Segbot : a multipurpose robotic platform for multi-floor navigation
(2014-12) Unwala, Ali Ishaq; Stone, Peter, 1971-
The goal of this work is to describe a robotics platform called the Building Wide Intelligence Segbot (segbot). The segbot is a two wheeled robot that can robustly navigate our building, perform obstacle avoidance, and reason about the world. This work has two main goals. First we introduce the segbot platform to anyone that may use it in the future. We begin by examining off-the-shelf components we used and how to build a robot that is able to navigate in a complex multi-floor building environment with moving obstacles. Then we explain the software from a top down viewpoint, with a three layer abstraction model for segmenting code on any robotics platform. The second part of this document describes current work on the segbot platform, which is able to non-robustly take requests for coffee and navigate to a coffee shop while having to move across multiple floors in a building. My contribution to this work is building an infrastructure for multi-floor navigation. The multi-floor infrastructure built is non-robust but has helped identify several issues that will need to be tackled in future iterations of the segbot.