자료유형 | 학위논문 |
---|---|
서명/저자사항 | Faster Than Real-Time GPGPU Radiation Pressure Modeling Methods. |
개인저자 | Kenneally, P. W. |
단체저자명 | University of Colorado at Boulder. Aerospace Engineering. |
발행사항 | [S.l.]: University of Colorado at Boulder., 2019. |
발행사항 | Ann Arbor: ProQuest Dissertations & Theses, 2019. |
형태사항 | 196 p. |
기본자료 저록 | Dissertations Abstracts International 81-04B. Dissertation Abstract International |
ISBN | 9781088318232 |
학위논문주기 | Thesis (Ph.D.)--University of Colorado at Boulder, 2019. |
일반주기 |
Source: Dissertations Abstracts International, Volume: 81-04, Section: B.
Advisor: Schaub, Hanspeter. |
이용제한사항 | This item must not be sold to any third party vendors. |
요약 | Solar radiation pressure (SRP) is a significant contributing dynamic force on spacecraft in all orbit regimes. Predicting, accommodating, and either leveraging or canceling its effect, is paramount to effective orbit determination, maneuver and mission design. As a result spacecraft numerical simulation requires computational models which provide the facility to model SRP with sufficient accuracy. However, typically the computationally intense nature of performing high-fidelity SRP evaluations has limited such evaluations to being an offline computation which generates lookup data. Precomputation limits the ability for a spacecraft dynamic simulation to accommodate the myriad time varying changes which occur to the spacecraft state during a mission. In the past decade the computer graphics industry has driven the development of highly parallel graphics processing units (GPU) capable of performing many thousands of floating point operations per second. General purpose GPU programming (GPGPU) has been leveraged particularly in Engineering and the Sciences where the high computational power of parallel GPU hardware presents the opportunity for significant increases in the size and dimension of computational problems now manageable on personal computers. This dissertation presents two modeling approaches which take advantage of the GPGPU aspect of commodity GPU hardware. The first contribution is a modeling approach which utilizes the vector graphics application programming interface (API) Open Graphics Library (OpenGL) and the GPGPU computing API Open Computing Language to develop a high geometric fidelity SRP modeling approach. The OpenGL-CL modeling approach computes SRP induced force and torque across a detailed spacecraft mesh model. The method utilizes the OpenGL-OpenCL shared context to facilitate modeling data between the two APIs. The OpenGL render pipeline is manipulated to render the sun-frame projected surface of the spacecraft into OpenGL Texture data objects. A custom OpenCL parallel reduction kernel is developed which subsequently computes the SRP force and torque across the spacecraft rendered into the OpenGL Textures. The method presents faster than real time computation speeds while accommodating spacecraft meshes with many thousands of vertices, arbitrary articulated components and detailed spacecraft material optical parameters. The second contribution is a GPU based parallel ray tracing modeling approach which ex- hibits faster than real time evaluation speeds. Techniques and algorithms from the computer graphics discipline are used to develop and implement a method which computes SRP force and torque across a detailed spacecraft triangulated mesh model. Efficient data structures such as bounding volume hierarchy (BVH) acceleration support a minimization of computational burden by reducing the ray-surface intersection search space. Accurate ray reflections are computed for complex materials by applying a Quasi-Monte Carlo integration method and importance sampling. Complex material bidirectional reflectance distribution functions (BRDF) are implemented with as both, ideal mirror-like specular and Lambertian diffuse, and as microfacet BRDF models. Arbitrary spacecraft articulation are accommodated at run time with no appreciable reduction in computational speed. Both SRP models utilize the latent computing power of the GPU which is exists in the large majority of consumer grade personal computing systems. Further access to latent computing power is enabled by the development of a software simulation communication middleware called Black Lion (BL). The third contribution of this thesis is the description of a novel software architecture and the design principles applied to the development of the BL software. Black Lion enables the integration of multiple local or distributed heterogeneous applications never intended to run in a cooperative settings. It is shown that BL enables access to more powerful latent personal computing resources by creating a means to transparently facilitate distributed simulation across multiple simulation nodes and computers. Finally, this dissertation demonstrates the utility of both modeling methods by their applica- tions in two case studies. Firstly, the high-fidelity SRP effects are computed for an ongoing asteroid sample return mission. Agreement between the OpenGL-CL methods is demonstrated. Both SRP modeling approaches make significant use of pre and post launch engineering data. The utility of direct access to a model's physical parameters is demonstrated in an analysis of contributors to possible error between modeled and estimated SRP accelerations. Secondly, capability of fast computational speed paired with high geometric resolution, of both OpenGL-CL and ray tracing methods, is demonstrated. Each method is employed in the simulation and long-term propagation of realistic multi-layer insulation (MLI) debris object mesh models and the effect of departing from the typical flat-plate MLI model is investigated. |
일반주제명 | Aerospace engineering. |
언어 | 영어 |
바로가기 | ![]() |