Search Results (1 - 6 of 6 Results)

Sort By  
Sort Dir
 
Results per page  

Bettaieb, Luc AlexandreA Deep Learning Approach To Coarse Robot Localization
Master of Sciences (Engineering), Case Western Reserve University, 2017, EECS - Electrical Engineering
This thesis explores the use of deep learning for robot localization with applications in re-localizing a mislocalized robot. Seed values for a localization algorithm are assigned based on the interpretation of images. A deep neural network was trained on images acquired in and associated with named regions. In application, the neural net was used to recognize a region based on camera input. By recognizing regions from the camera, the robot can be localized grossly, and subsequently refined with existing techniques. Explorations into different deep neural network topologies and solver types are discussed. A process for gathering training data, training the classifier, and deployment through a robot operating system (ROS) package is provided.

Committee:

Wyatt Newman (Advisor); Murat Cavusoglu (Committee Member); Gregory Lee (Committee Member)

Subjects:

Computer Science; Electrical Engineering; Robotics

Keywords:

robotics; localization; deep learning; neural networks; machine learning; state estimation; robots; robot; robot operating system; ROS; AMCL; monte carlo localization; particle filter; ConvNets; convolutional neural networks

Wolfe, Sage M.Heavy Truck Modeling and Estimation for Vehicle-to-Vehicle Collision Avoidance Systems
Doctor of Philosophy, The Ohio State University, 2014, Mechanical Engineering
This dissertation details the development of a state and position estimator for articulated heavy trucks based entirely on freely available on-board signals. The estimator consists of a quasi-linear vehicle dynamics model, tire cornering stiffness estimator, Kalman filter, and position integrator. Results from testing show that the estimator can provide lane-level (1.5 m) positioning accuracy in urban environments for the duration of typical GPS outages. A hybrid kinematic-dynamic model allows estimation of hitch angle to within half of a degree over the practical range of articulation angles. This presents novel contributions to the state of the art of trailer tire cornering stiffness estimation and hitch angle estimation. Government research has estimated that vehicle-to-vehicle (V2V) collision avoidance systems can address 72% of heavy truck crashes, but this requires localization of the truck and trailer in a variety of environments. Studies have shown that GPS cannot be reliably used for V2V in urban and some suburban environments. This estimator offers a potential supplement to GPS for V2V systems in these environments. Moreover, the current V2V messaging framework does not include an estimate of hitch angle. This can lead to missed warnings and false positives when the implicit assumption of zero hitch angle is grossly violated, such as turning at an intersection. Results from this research indicate that a reliable estimate can be provided without the addition of new sensors.

Committee:

Dennis Guenther (Advisor); Gary Heydinger (Committee Member); Junmin Wang (Committee Member); Anthony Luscher (Committee Member)

Subjects:

Automotive Engineering; Mechanical Engineering

Keywords:

vehicle dynamics; bicycle model; parameter estimation; state estimation; hitch angle; articulation angle; dead reckoning; V2V; vehicle-to-vehicle; collision avoidance; lateral dynamics; tire cornering stiffness; trailer

Brahma, AvraMethodologies for modeling and feedback control of the nox-BSFC trade-off in high-speed, common-rail, direct-injection diesel engines
Doctor of Philosophy, The Ohio State University, 2005, Mechanical Engineering
Over the past decade, modern technologies such as Direct Injection (DI), Exhaust Gas Recirculation (EGR), Variable Geometry Turbocharging (VGT), and most recently, High Pressure Common Rail (HPCR) fuel injection have narrowed the gap between Diesel engines and Spark-Ignition (SI) engines in terms of environmental impact. These improvements in Diesel engine technology are accompanied by several challenges. The modern diesel engine is a complex nonlinear system that must be controlled optimally to ensure that it meets the environmental regulations while maintaining its performance. One-loop-at-a-time tuning is no longer effective due to the complexity of the system. Currently, a significant part of time is spent in the optimization of engine performance, a luxury industry cannot afford in the increasingly competitive scenario. For this reason, the automotive industry is realizing the significance of model based and multivariate control. Fuel path control has mostly been of a feedforward nature due to lack of appropriate sensors. With sensors for emissions such as NOX beginning to become commercially viable, a need has arisen to formulate control paradigms that incorporate the emission optimization problem into the feedback control framework. Methodologies for modeling and feedback control of the NOX-BSFC trade-off are explored. Two different types of subsystems of a common high-speed Diesel engine are modeled based on the chosen input parameters. Linear models for the open-loop torque and NOX dynamics are proposed for each subsystem based on models published in the literature. These models are identified and demonstrated to be capable of reproducing essential system properties. A generic control paradigm is proposed that enables the explicit incorporation of the trade-offs between different output variables directly within the control framework. Two different instances of a constrained NOX-BSFC trade-off are demonstrated to exist in a Diesel engine. The first example relates to the existence of a NOX-BSFC trade-off in the context of torque control in the fuel path. The second example demonstrates the existence of a similar trade-off in the context of air flow control. The above control paradigm is applied to each example for the feedback control of the NOX-BSFC trade-off.

Committee:

Giorgio Rizzoni (Advisor)

Keywords:

Control; Pareto-optimal; Trade-off; Diesel; Common-rail; NOx; BSFC; Torque; Modeling; Sliding mode; State estimation

Gallagher, Jonathan G.Likelihood as a Method of Multi Sensor Data Fusion for Target Tracking
Master of Science, The Ohio State University, 2009, Electrical and Computer Engineering
This thesis addresses the problem of detecting and tracking objects in a scene, using a distributed set of sensing devices in different locations, and in general use a mix of different sensing modalities. The goal is to combine data in an efficient but statistically principled way to realize optimal or near-optimal detection and tracking performance. Using the Bayesian framework of measurement likelihood, sensor data can be combined in a rigorous manner to produce a concise summary of knowledge of a target’s location in the state-space. This framework allows sensor data to be fused across time, space and sensor modality. When target motion and sensor measurements are modeled correctly, these “likelihood maps” are optimal combinations of sensor data. By combining all data without thresholding for detections, targets with low signal to noise ratio (SNR) can be detected where standard detection algorithms may fail. For estimating the location of multiple targets, the likelihood ratio is used to provide a sub-optimal but useful representation of knowledge of the state space. As the calculation cost of computing likelihood or likelihood ratio maps over the entire state space is prohibitively high for most practical applications, an approximation computed in a distributed fashion is proposed and analyzed. This distributed method is tested in simulation for multiple sensor modalities, displaying cases where it is and is not a good approximation of central calculation. Detection and tracking examples using measured data from multi-modal sensors (Radar, EO, Seismic) are also presented.

Committee:

Randolph Moses (Advisor); Emre Ertin (Advisor); Lee Potter (Committee Member)

Subjects:

Electrical Engineering

Keywords:

target tracking; state estimation; distributed calculation; likelihood maps; likelihood ratio; data fusion; sensor fusion; sensor networks

Arlinghaus, Mark C.Autopilot Development for an RC Helicopter
Master of Science in Engineering (MSEgr), Wright State University, 2009, Electrical Engineering
The development of an autopilot system for an RC helicopter presents interesting challengesfrom both a hardware and controls standpoint. The system detailed in this thesis utilizes a 13 state Extended Kalman Filter (EKF) to fuse sensor data and provide a position/velocity/attitude estimate. A novel, state of the art hybrid PID/LQR controller is developed and compared with a full state Linear Quadratic Regulator (LQR). The hybrid controller uses a proportional position, PID velocity outer loop coupled with an inner loop LQR for attitude control. The entire system is developed and implemented in hardware to produce a functional autopilot. The unit was installed on an Align Trex 600 RC helicopter and demonstrated its ability to hover the helicopter at a desired location. Preliminary investigations have also showed the controller is capable of flying waypoints.

Committee:

Lang Hong, PhD (Committee Chair); Kefu Xue, PhD (Committee Member); Xiaodong Zhang, PhD (Committee Member)

Subjects:

Electrical Engineering; Engineering

Keywords:

Autopilot; UAV; EKF; LQR; PID; Helicopter; Unmanned; Control; State Estimation;VTOL

Horning, MarcusFeedback Control for Maximizing Combustion Efficiency of a Combustion Burner System
Master of Science in Engineering, University of Akron, 2016, Electrical Engineering
An observer-controller pair was designed to regulate the fuel flow rate and the flue-gas oxygen ratio of a combustion boiler. The structure of the observer was a proportional-integral state estimator. The designed controller was composed of a combination of two common controller structures: state-feedback with reference tracking and proportional-integral-derivative(PID). A discrete-time, linear state-space model of the combustion system was developed such that the linear controller and observer could be designed. This required establishing separate models pertaining to the combustion process, actuators, and sensors. The complete model of the combustion system incorporated all three models. The combustion model, which related the flue-gas oxygen ratio to the fuel and oxygen flow rates, was obtained using the mathematical formulas corresponding to combustion of natural gas. The actuators were modeled using measured fuel and oxygen flow rate data for various actuator signals, and fitting the data to a parametric model. The established nonlinear models for the combustion process and actuators required linearization about a specified operating point. The sensors model was then obtained using the predictive error identification technique based on batch input-output data. For the acquired model of the combustion system, a linear quadratic regulator was used to calculate the optimal state feedback gain. The classical controller gains were determined by tuning the gains and evaluating the simulation of the closed-loop response. Computer-aided simulations provided evidence that the controller and state estimator could regulate the desired set point in the presence of moderate disturbances. The observer-controller pair was implemented and verified on an experimental boiler system by means of an embedded system. Even in the presence of a disturbance resulting from a 50% blockage of the surface area of the air intake duct, the closed-loop system was capable of regulating the desired set point for slow-varying reference signal changes.

Committee:

Nathan Ida, Dr. (Advisor); Robert Veillette, Dr. (Committee Member); Kye-Shin Lee, Dr. (Committee Member)

Subjects:

Electrical Engineering; Engineering

Keywords:

PID control; combustion burner system; state-feedback control; combustion efficiency; kalman filter; PI state estimation; system identification; prediction error identification