This project deals with human robot-network collaboration for synchronization to desired reference velocities. A human interacts with the network of mobile robots by sending virtual source inputs to the leading robot. The information propagates through the network from each robot sensing its nearest neighbor. This research proposes a visual interface design to develop a real time, wireless communication channel between a human operator and the robot-network. The model developed introduces real time feedback from the human operator via a graphical interface of the relative positions of each robot in the network. The human operator’s ability to move the network cohesively with desired velocity trajectories require rapid information transfer, which is achieved using a delayed self-reinforcement (DSR) technique. We expect the human operator’s ability to move the network cohesively to improve with DSR hence enabling easier operation for the operator. The human-swarm interface designed has applications for semi-autonomous networks such as vehicle platoons. We can improve modern freight transportation safety and efficiency with a human remotely operating a robot-network of trucks.

This work describes an inexpensive and accurate gesture control implementation designed for an industrial setting. Sensing hand movements and being able to remotely operate devices without use of a tangible control can be useful, particularly in manufacturing applications where other methods of communication may not be available. One gesture recognition method is to use a camera or set of cameras to capture the motions of the user. However, this method imposes line-of-sight workspace constraints and is sensitive to environmental factors, such as consistent lighting conditions. My approach is to use an instrumented glove that detects the amount of bend in specific joints and sends those positions to a central processor that is programmed to recognize control gestures. Similar glove controllers are available but are either not well suited to an industrial setting because the sensors are vulnerable to metal dust and debris, or are not accurate enough to identify commands quickly and consistently. My version has custom sensors exactly fitted to this application and aims to have all sensors sealed and self contained to protect against contamination. This system is able to capture high resolution movement from the wearer and either save that data for machine training or send it immediately to be acted on. Going forward, onboard capabilities such as local gesture recognition will be added, as well as allowing the user to add custom gestures suited to their particular application.

Supraspinatus tendon tears are a type of rotator cuff tear, accounting for 15% of overhead workplace musculoskeletal injuries. These tears disproportionately affect blue-collar workers and cost millions in healthcare every year, but there is still relatively little known about the appropriate work-rest cycles to prevent the risk of occurrence during work. Directly measuring the rotator cuff in vivo is difficult because the supraspinatus is covered by the bursa sac, the acromion, and the deltoid, making its material properties hard to accurately record. This presents a need for a material that can model an in vivo shoulder tendon. There are many options of what materials can be used: organic and in vitro models are the most common, with relatively new inorganic models being designed. However, none of these models fulfill all modeling needs; overlap between all models is needed to get an idea of how an in vivo tendon accumulates damage. Organic models can provide tissue repair and degradation rates and these can be projected for a human supraspinatus. From in vitro studies stress-strain curves and maximum load can be recorded, and from inorganic models tear propagation can be observed. This work compiles research on candidates for tendon proxy materials by cross-referencing a variety of papers in tendon literature to find the foundational papers. Then builds off those with other works by the foundational authors or other highly regarded works that cite those foundational papers. From the collection of these papers, the shortcomings of current tendon modeling can accurately be seen, showing what research is needed to better model in vivo tendons. For instance, to confirm the hypothesized projection from organic models, psychophysical testing that isolates the supraspinatus needs to be conducted. Better modeling of tendons will allow for better prediction of appropriate work-rest cycles that may slow tendon fatigue damage.

Manufacturing workers are often subjected to many rigorous and repetitive shoulder and arm motions, usually leading to shoulder injuries. Assessing the likelihood of an injury before it occurs and adjusting practices accordingly can keep the individual from the severe pain that shoulder injuries can cause. This research focuses on creating such a predictive model to warn individuals before they sustain an injury. I created a mathmatical model to assess critical positions that would cause the most stress in the shoulder joint. A Kinect sensor locates the arm joints in space and my Matlab code calculates the expected reaction forces in the shoulder. My current results focus on single, static positions defined by common industry working positions. Future work will focus on dynamic positions and comparing the results from the mathematical model with biological indicators to determine if this predictive model is indicative of injury.