The majority of drones are extremely simple, their functions include flight and sometimes recording video and audio. While drone technology has continued to improve these functions, particularly flight, additional functions have not been added to mainstream drones. Although these basic functions serve as a good framework for drone designs, it is now time to extend off from this framework. With this Honors Thesis project, we introduce a new function intended to eventually become common to drones. This feature is a grasping mechanism that is capable of perching on branches and carrying loads within the weight limit. This concept stems from the natural behavior of many kinds of insects. It paves the way for drones to further imitate the natural design of flying creatures. Additionally, it serves to advocate for dynamic drone frames, or morphing drone frames, to become more common practice in drone designs.

The research presented in this Honors Thesis provides development in machine learning models which predict future states of a system with unknown dynamics, based on observations of the system. Two case studies are presented for (1) a non-conservative pendulum and (2) a differential game dictating a two-car uncontrolled intersection scenario. In the paper we investigate how learning architectures can be manipulated for problem specific geometry. The result of this research provides that these problem specific models are valuable for accurate learning and predicting the dynamics of physics systems.<br/><br/>In order to properly model the physics of a real pendulum, modifications were made to a prior architecture which was sufficient in modeling an ideal pendulum. The necessary modifications to the previous network [13] were problem specific and not transferrable to all other non-conservative physics scenarios. The modified architecture successfully models real pendulum dynamics. This case study provides a basis for future research in augmenting the symplectic gradient of a Hamiltonian energy function to provide a generalized, non-conservative physics model.<br/><br/>A problem specific architecture was also utilized to create an accurate model for the two-car intersection case. The Costate Network proved to be an improvement from the previously used Value Network [17]. Note that this comparison is applied lightly due to slight implementation differences. The development of the Costate Network provides a basis for using characteristics to decompose functions and create a simplified learning problem.<br/><br/>This paper is successful in creating new opportunities to develop physics models, in which the sample cases should be used as a guide for modeling other real and pseudo physics. Although the focused models in this paper are not generalizable, it is important to note that these cases provide direction for future research.

Drylands, though one of the largest biomes, are also one of the most understudied biomes on the planet. This leaves scientists with limited understanding of unique life forms that have adapted to live in these arid environments. One such life form is the hypolithic microbial community; these are autotrophic cyanobacteria colonies that can be found on the underside of translucent rocks in deserts. With the light that filters through the rock above them, the microbes can photosynthesize and fix carbon from the atmosphere into the soil. In this study I looked at hypolith-like rock distribution in the Namib Desert by using image recognition software. I trained a Mask R-CNN network to detect quartz rock in images from the Gobabeb site. When the method was analyzed using the entire data set, the distribution of rock sizes between the manual annotations and the network predictions was not similar. When evaluating rock sizes smaller than 0.56 cm2 the method showed statistical significance in support of being a promising data collection method. With more training and corrective effort on the network, this method shows promise to be an accurate and novel way to collect data efficiently in dryland research.

When rover mission planners are laying out the path for their rover, they use a combination of stereo images and statistical and geological data in order to plot a course for the vehicle to follow for its mission. However, there is a lack of detailed images of the lunar surface that indicate the specific presence of hazards, such as craters, and the creation of such crater maps is time-consuming. There is also little known about how varying lighting conditions caused by the changing solar incidence angle affects perception as well. This paper addresses this issue by investigating how varying the incidence angle of the sun affects how well the human and AI can detect craters. It will also see how AI can accelerate the crater-mapping process, and how well it performs relative to a human annotating crater maps by hand. To accomplish this, several sets of images of the lunar surface were taken with varying incidence angles for the same spot and were annotated both by hand and by an AI. The results are observed, and then the AI performance was rated by calculating its resulting precision and recall, considering the human annotations as being the ground truth. It was found that there seems to be a maximum incidence angle for which detect rates are the highest, and that, at the moment, the AI’s detection of craters is poor, but it can be improved. With this, it can inform future and more expansive investigations into how lighting can affect the perception of hazards to rovers, as well as the role AI can play in creating these crater maps.