RHex is a cockroach-inspired hexapod robot capable of walking, running, and climbing. This thesis presents the development of a novel, three dimensional, bipedal running gait for RHex using only two actuated degrees of freedom (DOF), one per compliant leg. To the author's knowledge, there are no previous, two DOF bipeds capable of running. In this thesis it is experimentally demonstrated that only body pitch and leg angle sensing are required. The controller includes three levels of proportional derivative controls for balancing, forward speed and leg tracking, as well a leg trajectory generator and a means of forward speed estimation. With the addition of yaw angle feedback, high repeatability was obtained. Details of a platform upgrade are also presented. Finally, a simulation model of the robot was developed that, in conjunction with a genetic algorithm optimization used to tune the controller and gait parameters, produced a stable gait similar to those observed in experiment.

<p> The communications and electronic systems that comprise a distributed control architecture for a robotic manipulator tie the high level control and motion planning to the electromechanical components. Custom solutions to this problem can be expensive in terms of time, cost, and maintenance. The integration of commercial off the shelf (COTS) motion controllers, combined with a robust communication standard, offers the potential to reduce the costs and development times for new robots. This thesis demonstrates an implementation of this architecture using commercial controllers and the CANopen communications bus on two existing dexterous robots. Testing is conducted to quantify the single joint performance of these modules. Additionally, the implementation of the system on a second robot arm was conducted in order to test the flexibility of the system for use with different actuators and feedback.</p>

Virtual reality has been used widely as a computerized medium for training. In addition to visual and auditory feedback, the addition of haptic feedback to virtual environments envisions promising applications for human motor skill acquisition, such as rehabilitation and training of surgeons. Generally speaking, there exist two kinds of virtual training paradigms, virtual practice and haptic guidance. Compared to virtual practice, haptic guidance conveys more information to the trainee, some of which might not be realizable in the physical world, in order to improve the training efficacy over practice. In this thesis, the most general form of haptic guidance, shared control, is introduced. A shared controller dynamically intervenes, through an automatic feedback controller acting upon the system, to modify the coupled system dynamics during training. The coupled system dynamics are to be selected to help expedite learning of the task. Specifically, an error-reducing shared controller is proposed and implemented for this dissertation. A series of human-participant studies have been conducted to test the efficacy of the proposed shared control approach for performance enhancement and training in virtual environments. During the experiments, performance is measured utilizing a Fitts'-like under-actuated target-hitting task. Experimental results indicate that an error-reducing shared control paradigm with fixed control gains does enhance performance of the participants during the target-hitting task when assistance is on. However, a one-month long training experiment reveals that the error-reducing shared control paradigm with fixed control gains exhibits negative efficacy during training in virtual environments. It is suggested that the fixed control gains implemented in the shared controller are one of the primary reasons for negative efficacy of haptic guidance for training. Therefore, a shared controller with performance-based gain adaptation was also tested. In addition to comparisons with the fixed-gain shared controller, this paradigm is also compared to virtual fixtures, another commonly used form of haptic guidance. The experimental results indicate that the performance-based progressive shared control paradigm results in significantly better training performance than all other haptic guidance training protocols implemented in this study. Hence, the training effectiveness of performance-based error reducing shared control is verified. Finally, this study is aimed to improve the design of progressive shared control algorithms for manual control tasks. Therefore, a perception study has been conducted quantifying human sensitivity to varying system dynamics. The results of this study form the basis for properly choosing the step size of the progressive control gains of shared controllers.

Motion planning research has been successful in developing planning algorithms which are effective for solving problems with complicated geometric and kinematic constraints. Various applications in robotics and in other fields demand additional physical realism. Some progress has been made for non-holonomic systems. However systems with complex dynamics, significant drift, underactuation and discrete system changes remain challenging for existing planning techniques particularly as the dimensionality of the state space increases. This thesis develops a novel motion planning technique for the solution of problems with these challenging characteristics. The novel approach is called Path Directed Subdivision Tree Exploration algorithm (PDST-EXPLORE) and is based on sampling-based motion planning and subdivision methods. PDST-EXPLORE demonstrates how to link a planner with a physical simulator using the latter as a black box, to generate realistic solution paths for complex systems. The thesis contains experimental results with examples with simplified physics including a second order differential drive robot and a game which exemplifies characteristics of dynamical systems which are difficult for planning. The thesis also contains experimental results for systems with simulated physics, namely a weight lifting robot and a car. Both systems have a degree of physical realism which could not be incorporated into planning before. The new planner is finally shown to be probabilistically complete.

An Electron Microscopy Proteomic Organellar Preparation (EMPOP) robot was developed as a tool for high-throughput preparation of subcellular fraction samples for electron microscopic identification. It will provide a means for validation of subcellular sample purity and confirmation of protein localization needed for organellar proteomics. / The device automates all chemical and mechanical manipulations required to prepare organelles for electron microscopic examination. It has a modular, integrated design that supports automated filtration, chemical processing, delivery and embedding of up to 96 subcellular fraction samples in parallel. Subcellular fraction specimens are extremely fragile. Consequently, the system was designed as a single unit to minimize mechanical stress on the samples by integrating a core mechanism, composed of four modular plates, and seven support subsystems for: (1) cooling, (2-3) fluid handling, (4-7) positioning. Furthermore, control software was developed specifically for the system to provide standardized, reproducible sample processing while maintaining flexibility for adjustment and recall of operational parameters. / Development of the automated process progressed from initial validation experiments and process screening to define operational parameters for preservation of sample integrity and establish a basic starting point for successful sample preparation. A series of successive modifications to seal the local environment of the samples and minimize the effect of fluidic perturbations further increased process performance. Subsequent testing of the robot's full sample preparation capacity used these refinements to generate 96 samples in approximately 16 hours; reducing the time and labor requirement of equivalent manual preparation by up to 1,000 fold. / These results provide a basis for a structured approach toward process optimization and subsequent utilization the device for massive, parallel preparation of subcellular fraction samples for electron microscopic screening and quantitative analysis of subcellular and protein targets necessary for high-throughput proteomics.

It is Zeus' anathema on biomedical engineering that we should agonise between the Scylla of the simulation of the physiological and anatomical findings and the Charybdis of the efficient implementation of a system. Previous simulations on the addition of the perceptual velocity tracking of a target to a biomimetic controller developed by Galiana allow the implementation of a novel oculomotor system (OCS) in a hardware device comprising two cameras and a neck, each of which able to produce yaw and pitch movement. The robotic system uses visual inputs and a variety of movements in order to track targets, namely slow pursuit (slow phase), saccades (fast phase), conjugate and vergence. The addition of prognosis and memory to the controller, leads to the need for new tactics to be presented that address the problem of fixating a target more efficiently. Since the length and the flatness of the pursuit bandwidth of the new controller are essential for the minimization of the need for saccades, a greater harmonization among the parameters of the controller is basic. The resulting OCS can now be used to label with a significance factor each target presented on the visual field and track the one which is the most conspicuous. The intensity and the speed of the target are the criteria used to select a target and to alternate the alertness state of the robot.

The goal of this research is to explore the control of a robotic arm in a visually guided reaching operation. Considering the role of the superior colliculus and the spinal cord in human arm movements, this study proposes simple approaches for the control of arm reaching. An efficient use of distance error signals between actual and desired end-point positions leads to reaching trajectories in two dimensions that are very similar to more traditional kinematic strategies. This implies that the nature of the premotor signal in the spinal cord may be related to simple error signals rather than trajectories. / The main conclusions to be stated from this work have two implications. First that a very simple biomimetic strategy can be used to control a robotic arm, and produces trajectories very similar to those seen in human reaching. Second, the hardware tests act as indirect proof that many concepts in biological motor control need to be revisited. For example, the need for complex kinematic computations is now de-emphasized.

Robot tasks are becoming more complex, as they can now be coded off-line by integrating the robot mathematical model into the controller. One of the main technological barriers in the robotics industry has been the reduction of discrepancies between the robot mathematical model and the actual robot geometry, by means of a process called geometric calibration. In this work, an inexpensive, time-effective and versatile closed-loop calibration method for serial robots is introduced. Contrary to most calibration methods, the proposed method does not require any external measurement system. The main idea is to fix the end-effector of a seven-degree-of-freedom 3D kinematic chain, formed with the robot joints and, if required, additional passive joints. Then, while performing a self-motion, i.e. changing the posture of the chain without changing its last link pose, the robot joint positions are recorded. Finally, the last link pose, computed using the mathematical model of the chain for the different postures, is subtracted from all others, thereby resulting in an overdetermined non-linear system of equations, from which the robot geometric parameters can be found. Two parameter representations were first chosen and applied to the CUATRO Arm, a serial four-degree-of-freedom redundant positioning robot, and the Thermo CRS A465, a seven-degree-of-freedom industrial robot. A versatile calibration set-up was then designed and manufactured to fix the end-effector of these robots. Moreover, different numerical identification methods were proposed and tested by means of simulation. Finally, the calibration method adopted here was applied to the CUATRO Arm and the Thermo CRS A465 robot. Although the experimental results correlated with the simulation results, the proposed calibration method was found to be highly sensitive to input noise.

<p>Given a desired function for an effector, what is its appropriate shape? This thesis addresses the problem of designing the shape of a rigid end effector to perform a given manipulation task. It presents three main contributions: First, it describes the contact kinematics of an effector as the product of both its shape and its motion, and assumes a fixed motion model to explore the role of shape in satisfying a certain manipulation task. Second, it formulates that manipulation task as a set of constraints on the geometry of contact between the effector and the world. Third, it develops tools to transform those contact constraints into an effector shape for general 1-DOF planar mechanisms and general 1-DOF spatial mechanisms, and discusses the generalization to mechanisms with more than one degree of freedom. </p><p> We describe the case studies of designing grippers with invariant grasp geometry, grippers with improved grasp stability, and grippers with extended grasp versatility. We further showcase the techniques with the design of the fingers of the MLab hand, a three-fingered gripper actuated with a single motor, capable of exerting any combination of geometrically correct enveloping or fingertip grasps of spherical, cylindrical, and prismatic objects of varying size. </p>

One important area for application of space robotics is autonomous on-orbit servicing of failed or failing spacecraft. An important aspect of these operations is the autonomous capture of the client satellite based on information obtained from a vision system. In this work, we describe laboratory experiments that verify the feasibility of autonomous capture of a slowly spinning non-cooperative satellite by a redundant serial manipulator. The main autonomous capture problem is divided in two separate tasks: the generation of a Cartesian trajectory to achieve the capture and the control of a manipulator to realize the generated Cartesian trajectory. Strategies that utilize the redundancy of a manipulator to optimize its posture are analyzed, implemented and used in the experimental validation of the autonomous capture. An online vision-based trajectory generation algorithm that generates a task-space velocity command to safely approach the target satellite and match its motion has been developed. The redundancy resolution and the trajectory generation algorithms are implemented and tested on two seven degree-of-freedom redundant manipulator systems: one located at McGill University and the other at the Canadian Space Agency. At both facilities, a scenario emulating the capture of a free-floating satellite have been created and used to validate the efficiency of the developed capture algorithm.