Search for: All records

In this work, we present three vision-based within-hand manipulation methods for a variable friction hand. The system is able to plan and execute actions such as in-hand sliding and large object rotation, which are required for many within-hand manipulation tasks. We present an experimental study using objects with various geometries, and show that all the methods can reliably achieve given target object poses. We discuss the advantages and disadvantages of the proposed methods with respect to positioning accuracy and efficiency. 

Deployment of sensors in hard-to-access locations can improve data gathering for scientific studies. We have developed a sensor emplacement system that can be mounted to unmanned aircraft systems with vertical takeoff and landing capabilities to autonomously auger a sensor into the ground. Various techniques can be chosen to enhance the augering process when certain characteristics of the soil are known. Moisture content and compressive strength are the soil characteristics that most impact the augering process, yet directly measuring them would require additional sensors to an already-burdened airframe. We address this through a novel means of predicting these soil characteristics within the first 30 s of an average 85 s augering evolution using onboard sensors and a Gaussian process regression scheme that predicts the soil moisture content and compressive strength with accuracy of 86.53% and 90.53% of the respective measured values. 

Deployment of sensors in hard-to-access locations can improve data gathering for scientific studies. We have developed a sensor emplacement system that can be mounted to unmanned aircraft systems with vertical takeoff and landing capabilities to autonomously auger a sensor into the ground. Various techniques can be chosen to enhance the augering process when certain characteristics of the soil are known. Moisture content and compressive strength are the soil characteristics that most impact the augering process, yet directly measuring them would require additional sensors to an already-burdened airframe. We address this through a novel means of predicting these soil characteristics within the first 30 s of an average 85 s augering evolution using onboard sensors and a Gaussian process regression scheme that predicts the soil moisture content and compressive strength with accuracy of 86.53% and 90.53% of the respective measured values. 

We design a compliant delta manipulator using 3D-printing and soft materials. Our design is different from the traditionally rigid delta robots as it is more accessible through low-cost 3D-printing, and can interact safely with its surroundings due to compliance. This work focuses on parallelogram links which are a key component of the delta robot design. We characterize these links over twelve dimensional parameters, such as beam and hinge thickness, and two material stiffness settings by displacing them, and observing the resulting forces and rotation angles. The parallelogram links are then integrated into a delta robot structure to test for delta mechanism behavior, which keeps the end-effector parallel to the base of the robot. We observed that using compliant hinges resulted in near-delta behavior, laying the groundwork for fabricating and utilizing 3D-printed compliant delta manipulators. 

Multirotor systems have traditionally been employed for missions that ensure minimal contact with the objects in their vicinity. However, their agile flight dynamics lets them sense, plan and react rapidly, and therefore perform highly dynamic missions. In this work, we push their operational envelope further by developing a complete framework that allows a multirotor to dock with a moving platform. Our approach builds on state-of-the-art and optimal methods for estimating and predicting the state of the moving platform, as well as for generating interception trajectories for the docking multirotor. Through a total of 25 field tests outdoors, we demonstrate the capabilities of our system in docking with a platform moving at different speeds and in various operating conditions. We also evaluate the quality of our system’s trajectory following at speeds over 2 m/s to effect docking within 10 s.