Activity
Time:
Ph.D. Dissertation Defense: Adam Plowcha
Date:
10:00 am –
11:00 am
Schorr Center
Room: 211
1100 T St
Lincoln NE 68588
Lincoln NE 68588
Additional Info: SHOR
“Autonomous, In-Ground, Sensor Emplacement Using Unmanned Aircraft Systems”
Close-proximity sensing of scientific phenomena improves data collection and quality. However, placing sensors near a data source in the field may be costly, pose hazards to personnel, disturb the surrounding or target environment, or corrupt the data the sensor is to collect. While remote sensing can overcome many of these challenges, sometimes it is absolutely required that a sensor be physically placed near the respective data source. Unmanned aircraft systems (UAS) are well-suited to dropping sensors in the most remote of locations. However, sometimes dropping a sensor in close proximity to a data source is not enough. For instance, seismic or moisture sensors function most effectively when placed into the ground. This thesis examines our solution of a UAS-mounted sensor emplacement system that can autonomously auger a sensor into the soil from two perspectives. First, as a multirotor UAS is limited in range compared to its fixed-wing counterparts, we must transport our system to the location where the sensor is to be placed. This involves the hardware and software necessary to deploy our system via parachute to its location. Then, our system must autonomously auger our sensor into the soil. Using emplacement systems with differing capabilities, we examine the hardware, software, and algorithms necessary to predict a failed augering attempt, estimate the soil composition in terms of moisture content and compressivity, and adapt the augering technique based on those estimated soil parameters. For failure prediction we implement a Markov Decision Process-based policy that can determine success or failure within 20 seconds of starting to auger. To determine soil conditions, we use a Gaussian Process Regression scheme that estimates soil moisture and compressive strength to within approximately 87 and 91 percent of their respective actual values. We then use the estimated soil conditions in concert with system performance parameters to drive an adaptive augering algorithm that succeeds 75 percent of the time in soil conditions that result in failure when using a constant rate drilling approach.
Committee Members:
Dr. Detweiler
Dr. Bradley
Dr. Roohi
Dr. Franz
Zoom Meeting: https://unl.zoom.us/j/94271346657
Close-proximity sensing of scientific phenomena improves data collection and quality. However, placing sensors near a data source in the field may be costly, pose hazards to personnel, disturb the surrounding or target environment, or corrupt the data the sensor is to collect. While remote sensing can overcome many of these challenges, sometimes it is absolutely required that a sensor be physically placed near the respective data source. Unmanned aircraft systems (UAS) are well-suited to dropping sensors in the most remote of locations. However, sometimes dropping a sensor in close proximity to a data source is not enough. For instance, seismic or moisture sensors function most effectively when placed into the ground. This thesis examines our solution of a UAS-mounted sensor emplacement system that can autonomously auger a sensor into the soil from two perspectives. First, as a multirotor UAS is limited in range compared to its fixed-wing counterparts, we must transport our system to the location where the sensor is to be placed. This involves the hardware and software necessary to deploy our system via parachute to its location. Then, our system must autonomously auger our sensor into the soil. Using emplacement systems with differing capabilities, we examine the hardware, software, and algorithms necessary to predict a failed augering attempt, estimate the soil composition in terms of moisture content and compressivity, and adapt the augering technique based on those estimated soil parameters. For failure prediction we implement a Markov Decision Process-based policy that can determine success or failure within 20 seconds of starting to auger. To determine soil conditions, we use a Gaussian Process Regression scheme that estimates soil moisture and compressive strength to within approximately 87 and 91 percent of their respective actual values. We then use the estimated soil conditions in concert with system performance parameters to drive an adaptive augering algorithm that succeeds 75 percent of the time in soil conditions that result in failure when using a constant rate drilling approach.
Committee Members:
Dr. Detweiler
Dr. Bradley
Dr. Roohi
Dr. Franz
Zoom Meeting: https://unl.zoom.us/j/94271346657