Browsing by Subject "Computer Science and Engineering"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Autonomous altitude estimation of a miniature helicopter using a single onboard camera.
    (2010-04) Cherian, Anoop
    Autonomous estimation of the altitude of an Unmanned Aerial Vehicle (UAV) is extremely important when dealing with flight maneuvers like landing, steady flight, etc. Vision based techniques for solving this problem have been underutilized. In this thesis, we propose a new algorithm to estimate the altitude of a UAV from top-down aerial images taken from a single on-board camera. We use a semi-supervised machine learning approach to solve the problem. The basic idea of our technique is to learn the mapping between the texture information contained in an image to a possible altitude value. We learn an over complete sparse basis set from a corpus of unlabeled images capturing the texture variations. This is followed by regression of this basis set against a training set of altitudes. Finally, a spatio-temporal Markov Random Field is modeled over the altitudes in test images, which is maximized over the posterior distribution using the MAP estimate by solving a quadratic optimization problem with L1 regularity constraints. The method is evaluated in a laboratory setting with a real helicopter and is found to provide promising results with sufficiently fast turnaround time.
  • Thumbnail Image
    Item
    Path planning algorithms for robots in a data muling system
    (2010-12) Bhadauria, Deepak
    We study two path planning problems that arise in data muling systems where robots are charged to collect data from wireless devices dispersed across a large environment. In such applications, deploying a network of stationary wireless sensors may be infeasible when many relay nodes are deployed to ensure connectivity. Instead, a few robots can be used as data mules to collect data from these devices. The first problem studied in this thesis is to find tours for multiple robots so as to collect data from all sensors in the least amount of time. We refer to this problem as the Data Gathering Problem (DGP). We assume that sensors have a uniform disk communication model. In this model, data can be downloaded from a sensor with fixed rate inside its communication disk. We present an optimal algorithm for the one dimensional version of DGP. For the two dimensional version we present a constant factor approximation algorithm. Afterwards, we present field experiments in which an autonomous robotic data mule collects data from sensor nodes deployed over a large area. Next, we study data collection problem with a more realistic communication model for sensors. In experiments we found that the time taken to download data from a sensor s is a function of the locations of the robot and s: If the robot is a distance rin away from s, it can download the sensor’s data reliably in Tin units of time. If the distance is greater than rin but less than rout , robot can still download data but due to higher packet loss probability the average download time Tout is higher (Tout > Tin). We refer to this model as the Two-Ring communication model and the corresponding path planning problem as the Two-Ring Tour (TRT) problem. We present a constant factor approximation algorithm for the general case. The algorithm uses a polynomial time approximation scheme as a subroutine. Though the scheme has polynomial running time, its running time is impractically large. It is also very complex to implement. Therefore we study special cases of the TRT problem and present efficient algorithms for them. For robotic data mules to be useful, the robots must be capable of operating in the field for extended periods of time. Therefore, in the last part of the thesis we initiate ii the study of solar energy harvesting for robotic navigation. Our primary contribution is an experimental model of energy consumption and harvesting as a function of environmental parameters. We demonstrate the utility of this model in a simple navigation task.
  • Thumbnail Image
    Item
    Towards Dynamics Modeling for an Autonomous Underwater Vehicle (AUV) in Experimental and Simulated Settings
    (2020) Orpen, Kevin
    Autonomous Underwater Vehicles (AUVs) have been in development in recent decades to address the difficulties and high costs of oceanic exploration, with a myriad of applications including marine life monitoring and search and rescue operations. An underwater robot in development by the Interactive Robotics and Vision (IRV) Laboratory at the University of Minnesota is LoCO, a Low Cost Open-Source AUV aiming to reduce the current high cost of entry into underwater robotics. One aspect critical to its capacity as an AUV is its autopilot system, which enables stability augmentation and predictable control behavior. Each new AUV comes with unique characteristics, requiring distinctive autopilot designs. This research seeks to prove the hypothesis that the known properties common to underwater environments (e.g., buoyancy and drag forces) can be characterized alongside parameterized variables adaptable to various AUV configurations. This understanding will lead to the efficient development of autopilot systems based on both dynamics modeling and experimental data, opposed to the purely experimental approximation of control parameters. Focusing on LoCO, this particular research centered on the development of a simulation program in Gazebo utilizing Robot Operating System (ROS) that has the potential to reduce time and cost spent on physical testing. Various physics aspects for simulated locomotion were considered alongside the implementation of initial underwater forces. Experimental data from physical testing was collected to characterize LoCO’s forward motion to aid in this initial modeling. Further evaluation and validation of dynamics modeling will build upon this framework, assisting in future control system development.