Smart Sensor Networks
Smart Sensor Networks (S-Nets) consist of a set of distributed devices which are capable of computation, communication and sensing. Our work addresses the creation of an information layer on top of the sensor nodes. The information processing issues include the representation of information and knowledge, the processing of that information, and the development of efficient, robust, scalable algorithms; we have contributed distributed algorithms for leadership protocols, coordinate frame and gradient calculation, reaction-diffusion pattern formation, and level set methods to compute shortest paths through the net.
Treadmills are an example of a locomotion device allowing a user to walk in a relatively normal manner without significant change in actual location. Our research deals with combining more sophisticated locomotion devices with visual displays in order to construct true locomotion interfaces which will allow a user to interact with a virtual world by walking through that world.
The goal of this project is to add a sense of contact and manipulation in the CAD design of mechanical assemblies. Part interaction, assembly, and manipulability can then be evaluated without fabrication of physical prototypes. A haptic device, the Sarcos Dextrous Arm Master, is being employed as a real-time interface to the Geometric Design and Computation (GDC) research group's Alpha_1 CAD/CAM system.
In order for humans and machines to interact naturally together, there needs to be a means for the machine to unobtrusively monitor what the human is doing, especially with the hands and fingers. Miniaturized optical sensors placed on the fingernail measure the blood perfusion in the fingertip, which changes as a function of finger posture and touching forces at the fingertip. This research aims to develop models of the fingertip and fingernail sensor system that can predict the state of the fingers based on the sensor signals. Once the models are accurately calibrated, the fingernail sensor system could be used to teleoperate a robot, interact with a virtual environment, or learn about how humans interact with the real world through tactile manipulation.
As robots become increasing complex and more like humans, they will need to have increasing numbers of muscles or actuators within a compact volume. The goal of this research is to develop new types of actuator systems, architectures, and methods of control that would allow arrays of hundreds or even thousands of small but powerful actuators to be arranged and controlled in a fast, efficient, and scalable manner. Currently, this is achieved by embedding Shape Memory Alloy "muscles" within a network of biologically inspired "robotic blood vessels" that fluidically distributes thermal energy to and from any actuators in the array using only a small number of valves.
Neural Control of Artificial Arms
Signals from electrodes placed in remnant peripheral nerves of an amputee are used to control an artificial arm. The electrodes are longitudinal intrafasicular electrodes (LIFEs) developed by Ken Horch of the Department of Bioengineering. The electrodes can be used to detect efferent neurons for control or to stimulate afferent neurons for sensation. Data from human subjects has been collected and is now being analyzed. Current work is focused on the determination of the statistical nature of the neural signals. The results will aid in the design of optimal signal processing of the neural signals for control use.
Stability of Quadruped Trotting using Directional Compliant Legs
Trotting quadrupeds such as dogs have compliant joints. It has been shown that the kinematic configuration of the fore and hind limbs creates a directional compliance with reference to the shoulder/hip joints. The term directional compliance is defined as the compliance tensor as seen by a joint or body. The off-diagonal terms of the tensor cause a directionally compliant spring to have different properties than simple springs. For instance, if a simple spring is compressed, it will have a normal force opposite the direction of the displacement. A directionally compliant spring when compressed in the same manner will not only have the normal force but also have a shear force. This type of compliance can provide dynamic stability by directing the ground reaction forces through the center of mass of the body. Dynamic simulations have shown that stability can be achieved passively with such as approach.
Compliant Frame Modular Robotic Systems
Mobile robots are integral to space exploration and many military activities. This research examines a new robot architecture that provides advanced maneuverability, suspension to adapt to uneven terrain, and modularity to adapt to different tasks. This is accomplished by coupling differentially steered axle modules with compliant frame elements that allow relative roll, pitch, and yaw between the axles. While this system provides advanced capabilities with a simple structure, it does present a number of research issues fundamental to its implementation which are our current focus: motion control and planning, dynamic stabilization, sensor instrumentation, and data fusion.
Miniature Climbing Robots with Under-Actuated and Hybrid Kinematic Structures
Climbing robots serve a multitude of functions ranging from military reconnaissance to industrial inspection and service. Miniature climbing robots, in particular, are well suited to operating in the confined environments typical of these applications, but the challenge for robots at these smaller scales is that their capabilities are limited by actuator capacity and space limitations. This research strives to improve the capability of miniature climbing robots using under-actuated mechanical structures that allow one motor to power multiple joints and in hybrid joint structures that allow a single joint to perform multiple functions. A key part of this development is advanced motion planning and control strategies that exploit the mechanical features.
Augmenting Haptic Interfaces
In everyday activities, we use our hands to assess the size, shape, hardness, and texture of objects; however, many of the sensations used to identify these properties are absent from current haptic interfaces. Augmenting haptic devices with tactile feedback creates the potential to tap into our innate experience when conducting manual interactions with remote or virtual environments. The focus of this research is to explore new avenues of tactile feedback used in combination with commercial haptic devices, and includes new device development and psychophysical evaluation. This research has applications in telemanipulation (e.g., surgical robotics) and virtual reality (e.g., medical training).