Author ORCID Identifier

https://orcid.org/0009-0004-3439-0892

Semester

Spring

Date of Graduation

2023

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Mechanical and Aerospace Engineering

Committee Chair

Guilherme Augusto Silva Pereira

Committee Member

Yu Gu

Committee Member

Mario Perhinschi

Abstract

This thesis presents a path optimization solution for a robot in two different constrained 3-dimensional (3D) environments. The robot is required to travel from its current position to a goal position following minimum cost paths (optimal paths). The first environment has 3D obstacles that interfere with the robot’s path. The path cost for this environment accounts for the minimum distance traveled by the robot from the start to the goal position while avoiding obstacles. The second environment is the atmosphere of Venus, specifically a flyable region of this atmosphere with characteristics similar to Earth’s. This environment has strong westward winds that require a more complex cost function. The path cost also accounts for energy expenditure, such as thrust or drag, and energy accumulation, such as charging using the robot’s solar panels and gains of potential energy (e.g., due to upward directional winds). In this case, we can add to the path cost function the localization cost of the robot. Localization is simulated in the environment by the use of cameras pointing to the surface of the planet, with yields lower localization error when the vehicle is close to the surface. The approach proposed in this paper uses genetic algorithms, a heuristic search that, based on a population of initially feasible paths and a set of biologically inspired operations, finds a low-cost path. Path feasibility is assured by computing local reachability regions based on different factors such as wind velocity, obstacles, and the maximum speed of the robot. The method is illustrated through a series of simulations that show our results as a function of the number of iterations and path population sizes. Finally, a comparison with different planners is made in order to show that the genetic algorithms allow for more efficient and easier implementations.

Download

Share

COinS