Stochastic control via chance constrained optimization and its application to unmanned aerial vehicles
Abstract/Contents
- Abstract
- Automation has had a strong presence in manufacturing for centuries, however its use in society has been very limited. The most ubiquitous robot in everyday life is the Roomba which has sold over 6 million, but the vacuuming task of this robot is simple in comparison with the difficult tasks performed in the manufacturing sector. Automation is successful in manufacturing because the task and environment can be well-defined to simplify the problem, whereas for automation to penetrate into society, it needs to safely operate in the presence of significantly greater uncertainty. Applications that could benefit from increased autonomy include: robot-assisted surgery, energy efficient control of buildings, autonomous control of vehicles, robotic assistance for elderly and disabled people, routing aircraft around weather, and home automation for tasks such as folding laundry, loading and emptying a dish washer, or tidying up a room. All of the above applications are examples of stochastic systems which contain three sources of uncertainty: process uncertainty, sensing uncertainty, and environment uncertainty. In order to safely operate in the intended domains, the system must account for all three types of uncertainty in generating a safe control strategy. This problem of generating control inputs for systems under uncertainty is commonly referred to as the stochastic control problem. One way of formulating this problem is as a chance constrained optimization problem that restricts the risk of violating the system's constraints to be below a user supplied threshold. This thesis develops several extensions over existing chance constrained programming solutions. The feedback controller is used to shape the uncertainty of the system to facilitate the satisfaction of the stochastic constraints, enabling previously infeasible solutions. Systems with component failures are also studied, and the computational complexity is drastically reduced over previous solution methods. The chance constrained framework is also extended to handle systems operating in uncertain environments. A novel hybrid approach is developed that uses a combination of sampling and analytic functions to represent the uncertainty. This approach results in a convex optimization program, guaranteeing the optimal solution and reducing the complexity over other methods. The formulation is further extended to incorporate future measurements of the uncertain environment to increase the performance of the system. The proposed stochastic control methods are solved in real-time to plan trajectories for a quadrotor unmanned aerial vehicle navigating in a three-dimensional cluttered, uncertain environment. The solution method enables the quadrotor to explore the environment to gather more information, allowing it to successfully complete its objective.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2012 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Vitus, Michael Peter | |
|---|---|---|
| Associated with | Stanford University, Department of Aeronautics and Astronautics | |
| Primary advisor | Tomlin, Claire J, 1969- | |
| Thesis advisor | Tomlin, Claire J, 1969- | |
| Thesis advisor | Boyd, Stephen P | |
| Thesis advisor | Rock, Stephen M | |
| Advisor | Boyd, Stephen P | |
| Advisor | Rock, Stephen M |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Michael P. Vitus. |
|---|---|
| Note | Submitted to the Department of Aeronautics and Astronautics. |
| Thesis | Ph.D. Stanford University 2012 |
| Location | electronic resource |
Access conditions
- Copyright
- © 2012 by Michael Peter Vitus
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...