<p> This thesis describes the design, development, and evaluation of a novel "self-sealing" suction technology for grasping. As humans desire robots capable of handling an increasingly diverse set of tasks, end effectors that are able to grasp the widest possible range of object shapes and sizes will be needed to achieve the desired versatility. Technologies enabling the exertion of local pulling contact forces (e.g. suction) can be extraordinarily useful toward this end by handling objects that do not have features smaller than the grasper, a challenge for traditional grippers. However, simple operation and cost effectiveness are also highly desirable. </p><p> To achieve these goals, we have developed a self-sealing suction technology for grasping. A small valve inside each suction cup nominally seals the suction port to maintain a vacuum within the system. Through the reaction forces of object contact, a lever action passively lifts the valve to engage suction on the object. Any cups not contacting the object remain sealed. In this way, a system with a large number of cups may effectively operate using any subset of its cups, even just one, to grasp an object. All cups may be connected to a central vacuum source without the need for local sensors or powered actuators for operation, forming a simple, compact, cost effective system. </p><p> This thesis begins with the detailed design and analysis of the self-sealing suction technology. An extensive evaluation of the technology's robustness and performance demonstrates its features and limits. This includes self-seal quality and leakage, object seal and reseal, cycle performance, and normal and shear force-displacement, among other characterizations. It then describes the development of several devices utilizing the technology. The potential impact of the technology is highlighted through applications of human-controlled, robotic, and aerial grasping and perching. Finally, mathematical tools are developed to analyze potential grasps developed using the technology. </p><p>

<p> The past decade has seen the rapid development and deployment of unmanned systems throughout the world in both civilian and military applications. Significant development has been led by the Department of Defense (DoD), which has sought to develop and field military systems, such as unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs), with elevated levels of autonomy to accomplish their mission with reduced funding and manpower. As their role increases, such systems must be able to adapt and learn, and make nondeterministic decisions. Current unmanned systems exhibit minimal autonomous behaviors. As their autonomy increases and their behaviors become more intelligent (adapting and learning from previous experiences), the state space for their behaviors becomes non deterministic or intractably complex. </p><p> Consequently, fielding such systems requires extensive testing and evaluation, as well as verification and validation to determine a system&rsquo;s performance and the acceptable level of risk to make it releasable &ndash; a challenging task. To address this, I apply a novel systems perspective to develop a model-based framework to predict future system performance based on the complexity of the operating environment using newly introduced complexity measures and learned costs. Herein I consider an autonomous military ground robot navigating in complex off-road environments. Using my model and data from Defense Advanced Research Projects Agency (DARPA)-led experiments, I demonstrate the accuracy with which my model can predict system performance and then validate my model against other experimental results.</p>

<p> Field mapping and estimation become a challenging problem, with their various applications on non-linear estimation, geolocation, and positioning systems. In this research, we develop novel algorithms based on interval analysis and introduce a solution for autonomous map construction, field mapping, geolocation, and simultaneous localization and mapping (SLAM), providing applications on indoor geolocation and other potential areas. </p><p> Generally, the localization algorithm includes a quasi-state estimation and a dynamic estimation. Quasi-static estimation collects each single measurement and give a group of estimation intervals on the pre-constructed field map. Results from quasi-static estimation are processed into the dynamic estimation algorithm, having properties of removing redundant intervals while keep the best estimation results. Sizes of estimation from quasi-static estimation are proved to be related to the resolution of the map and the quality of the sensor. Based on quasi-state estimation algorithm, we develop an algorithm to fuse different type of measurements and discuss the condition when this algorithm an be applied effectively. </p><p> Having theoretical guarantees, we apply these algorithms to augment the accuracy of cell phone geolocation by taking advantage of local variations of magnetic intensity. Thus, the sources of disturbances to magnetometer readings caused indoors are effectively used as beacons for localization. We construct a magnetic intensity map for an indoor environment by collecting magnetic field data over each floor tile. We then test the algorithms without position initialization and obtain indoor geolocation to within 2m while slowly walking over a complex path of 80 meters. The geolocation errors are smaller in the vicinity of large magnetic disturbances. After fusing the magnetometer measurement with inertial measurements on the cell phone, the algorithm yields even smaller geolocation errors of under 50cm for a moving user. </p><p> The map construction and geolocation algorithms are then extended to realize the SLAM, with hierarchical structure of estimation update and localization update. When a new user steps into a random map, the dead reckoning algorithm with assistance of IMU and Kalman filter provides initial estimation of position on the map, which coordinates the corresponding reading of magnetic field intensity as well as all other sources such as WiFi received signal strength (RSS), to construct an initial map. Based on the initial map, we then apply the localization algorithm to estimate new geolocations consequently and fuse the estimation intervals both from IMU and from crowd-sourced field maps to reduce the estimation size and eventually revise the map as well as the geolocation. </p><p> In this research, we have built up mathematical model and developed mathematical solutions with corresponding theories and proofs. Our theoretical results connect geolocation accuracy to combinations of sensor and map properties. </p>

<p> Unmanned aerial vehicles are quickly becoming a significant and permanent feature in today's world of aviation. Amongst the various types of UAVs, a popular type is the quadcopter. Also referred to as a quadrotor, this rotor craft's defining feature is that it has four propellers. While its use is common in the hobbyist community, this aircraft's use within industry is blooming. </p><p> Presented are the efforts to design and implement a controller for a BeagleBone based quadcopter. As part of this effort, characteristics of the quadcopter were experimentally determined. These characteristics consist of physical properties of the quadcopter, such as the moments of inertia, the motor performance characteristics, and variance within its sensors. A model was then created and implemented within a MATLAB environment to simulate the flight of the quadcopter. With a simulated environment created, a controller was designed to control the flight of the quadcopter and a Kalman Filter was implemented to filter a sensor input. These designs were then verified in the simulated environment.</p>

<p> Astronaut hand fatigue during Extravehicular Activity (EVA) and EVA training is a critical risk in human space exploration. Improved glove designs over the past forty years have reduced hand fatigue, but limitations of the technology prevent major improvements to reduce hand fatigue. Therefore, a mechanism to assist astronauts by reducing hand fatigue was explored. Many organizations have already developed exoskeletons to assist astronauts, but all mechanisms developed required electrically powered actuators and control systems to enhance grip strength. However, astronauts already possess the strength required to actuate the glove; what is needed is a method to reduce fatigue without introducing electromechanical complexity. A passive mechanical system was developed as a proof-of-concept to test the feasibility of an unpowered exoskeleton to maintain static grip around an object. The semi- rigid nature of an inflated pressure glove provided an ideal substrate to mount a mechanism and associated components to allow an astronaut to release his/her grip inside the glove while maintaining attitude, as the mechanism will keep the glove closed around an object.</p><p> Three prototypes were fabricated and tested to evaluate the architecture. The final two prototypes were tested on a real pressure suit glove at Final Frontier Design (FFD), and the third mechanism demonstrated attachment and basic operating principles. At University of California (UC) Davis, pressure glove analogs were fabricated from a baseball batting glove and polystyrene to simulate a real pressure glove without the risk of testing in a reduced pressure environment (i.e. a glove box). Testing of the third prototype showed a reduction in fatigue as measured by Maximum Voluntary Contraction (MVC) grip force over a 30 second period when the mechanism assisted gripping an object.</p>

<p> Paralysis or paresis induced by upper motor neuron damage often leaves the lower limbs dysfunctional for basic activities such as walking and climbing stairs. Nearly five and one half million people in the United States, or approximately one in fifty, have some degree of paralysis [4]. The sustained duration and high level of impairment attributed to paralysis motivates research and development for technologies that alleviate the associated deficiencies.</p><p> The Indego exoskeleton (Parker-Hannifin, OH) and Chimera muscle stimulator are mechatronic devices developed for the reanimation of paretic limbs. Indego employs electric motors to actuate an orthosis for the restoration of controlled legged mobility, while the Chimera interfaces with the nervous system through transcutaneous electrical stimulation to administer functional electrical stimulation (FES). Described herein are rehabilitative intervention methods that: 1) enable paraplegics to ascend and descend stairs with a lower limb exoskeleton; 2) enhance exoskeleton assisted walking with supplemental FES to overcome moderate to severe spasticity; and 3) suppress clonus using FES during seated mobility. Chapter 2 describes the hardware developed for and/or employed in this research.</p><p> Chapter 3 describes the development and assessment of a controller for the Indego that enables paraplegics to ascend and descend stairs. The stair controller expands on a previous implementation of predefined trajectory tracking with an emulated passive state that enables gravity to extend the leg until it meets the next stair tread, then a trajectory is calculated in real-time to perform the intended task of stair ascent independent of step height. The ascent and descent controllers were evaluated by three paraplegic users who traversed numerous size stairs safely within two hours of tuning and training. The resulting controller enabled stair climbing with light exertion despite complete paraplegia.</p><p> Subjects with moderate to severe spasticity are typically ineligible for exoskeleton assisted gait due to pathological muscle activation that opposes exoskeleton mediated motion. A novel supplemental stimulation controller was implemented with the Indego exoskeleton and integrated Chimera stimulator in an effort to expand the inclusion criteria of exoskeletons to individuals with severe spasticity whereby FES enhances the synergy between muscles and motors. Chapter 4 explores the effects of spasticity and FES on robot mediated gait for paraplegics and describes the hybrid system&rsquo;s controller that enabled two paraplegic individuals with moderate to severe spasticity to achieve substantially improved gait kinematics.</p><p> Mobility impairment of paraplegia can also entail clonus, a self-exciting reflex that can manifest as involuntary shaking of the ankle, a common pathology experienced during wheelchair propulsion. Chapter 5 expands the frontiers of clonus research with the first reported evaluation methods wheelchair clonus 2 and the efficacy of a novel FES intervention to treat pathological clonus during wheelchair propulsion over rough terrain. The clonus intervention was shown to robustly suppress clonus. The treatment may provide a noninvasive and economical alternative to invasive and commonplace pharmacological interventions. </p><p> The remainder of the introduction serves to provide background information pertaining to the nervous system, neurological impairment and the state of the technologies used to restore deficiencies that arise from comorbidities of paralysis.</p><p>

Thesis (Ph. D.)--Drexel University, 2007. / Includes abstract and vita. Includes bibliographical references (leaves 141-147).

<p> Technology in robotics has improved significantly in recent years. While the majority of research has focused on improving existing methods, it is advantageous to challenge these established methodologies and develop new solutions. This new research centers on a novel method of robot movement design. The proposed model concentrates on a robot containing three steerable wheels, allowing the mobile robot to reach the desired orientation and coordinates with minimal movement. This goal is accomplished by simultaneously moving and rotating the robot while moving in a straight path, unlike the movement provided by standard wheeled vehicles. This method provides greater control of performance and more power of movement on various surfaces, compared to using Omni wheels, which contains the design with the greatest similarity to this proposed method. While this new method may result in added complexity due to the goal-based flexible constraints in speed, wheel rotation, and overall movement, this complication may be mitigated by using appropriate software and hardware.</p><p>

<p> Prior work at Marquette University developed the Marquette Prosthesis, an active transtibial prosthesis that utilized a torsional spring and a four-bar mechanism. The controls for the Marquette Prosthesis implemented a finite state control algorithm to determine the state of gait of the amputee along with two lower level controllers, a PI moment controller to control the moment during stance and a PID position controller to control the position during stance. The Marquette Prosthesis was successful in mimicking the gait profile presented by Winter. However, after completing human subject testing, the Marquette Prosthesis was insufficient in trying to match the gait profile of those who varied from this textbook stride. </p><p> Active transtibial prostheses typically apply finite state control algorithms that struggle with cadence and gait variability of the amputee. Recent work in artificial neural networks (ANN) have shown the possibility to predict the user's intent which can be used as an input signal in an improved controller. The Marquette Prosthesis II was developed that uses a stiffness controller to control the relationship between the position and torque of the ankle. A model of the improved Marquette Prosthesis II was developed in Simulink to ensure that the stiffness controller was robust enough and that this type of control was possible with the limitations of the Marquette Prosthesis, i.e., the link lengths, torsional spring and motor. The mechanical system of the Marquette Prosthesis was then changed such that the spring was in series between the motor and four-bar mechanism to establish a relationship between the motor position, torque of the spring and four-bar mechanism. The control hardware was selected and the stiffness controller was implemented on the Marquette Prosthesis II. The Marquette Prosthesis II control algorithm was tested and validated to show that this approach is feasible.</p><p>

<p> Control of robot manipulators can be greatly improved with the use of velocity and torque feedforward control. However, the effectiveness of feedforward control greatly relies on the accuracy of the model. In this study, kinematics and dynamics analysis is performed on a six axis arm, a Delta2 robot, and a Delta3 robot. Velocity feedforward calculation is performed using the traditional means of using the kinematics solution for velocity. However, a neural network is used to model the torque feedforward equations. For each of these mechanisms, we first solve the forward and inverse kinematics transformations. We then derive a dynamic model. Later, unlike traditional methods of obtaining the dynamics parameters of the dynamics model, the dynamics model is used to infer dependencies between the input and output variables for neural network torque estimation. The neural network is trained with joint positions, velocities, and accelerations as inputs, and joint torques as outputs. After training is complete, the neural network is used to estimate the feedforward torque effort. Additionally, an investigation is done on the use of neural networks for deriving the inverse kinematics solution of a six axis arm. Although the neural network demonstrated outstanding ability to model complex mathematical equations, the inverse kinematics solution was not accurate enough for practical use.</p><p>