Reachability-based control for human and autonomous agents in adversarial games
Abstract/Contents
- Abstract
- Solutions to multi-agent adversarial games are useful in many applications where control actions must be taken in the presence of opposing agents, moving obstacles, or other external disturbances. These applications range from collision avoidance for autonomous cars and air traffic control to team coordination in the military and police. In addition to controlling robotic or autonomous agents, solutions to these games can play an essential role in providing guidance to humans, who may be participating either in supervisory roles or as agents themselves. The critical challenges to obtaining such solutions are the need to robustly account for adversarial actions in scenarios with many agents, and to do so in a computationally tractable fashion. In addition, to be helpful to humans the solutions must be presented so that they fit within the decision-making framework of the humans. This thesis develops solution methods for two types of related multi-agent adversarial games: reach-avoid games, where an attacking agent attempts to reach a target while avoiding other, defending agents, and pursuit-evasion games, where an evading agent attempts to avoid capture by pursuing agents. First, a Hamilton-Jacobi-Isaacs reachability formulation is used to compute guaranteed, optimal solutions to a reach-avoid game involving a single attacker and defender. Then an open-loop reachability formulation is introduced to allow solutions to be found in real time for games involving multiple agents on a side. An efficient method for computing open-loop safe-reachable sets is introduced and used to compute solutions for an attacking agent in reach-avoid games with multiple defenders, and for pursuing agents in pursuit-evasion games involving a single evader and multiple pursuers. These reachability-based game solutions are used to develop intuitive, automated tools to guide human agents in adversarial games as well as to control robotic assets like unmanned aerial vehicles (UAVs). The tools are validated using a novel testbed incorporating smartphones and UAVs in a system with human agents. These experiments demonstrate the use of the tools in realistic scenarios involving varied terrain and noisy communications. In these experiments, the human agents not only directly utilize the computed solutions, but are also able to modify the plans using the associated reachability information when communications failure and other factors render the assumptions made by the automation invalid.
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 | Huang, Haomiao |
|---|---|
| Associated with | Stanford University, Department of Aeronautics and Astronautics |
| Primary advisor | Tomlin, Claire J, 1969- |
| Thesis advisor | Tomlin, Claire J, 1969- |
| Thesis advisor | Klein, Daniel |
| Thesis advisor | Rock, Stephen M |
| Advisor | Klein, Daniel |
| Advisor | Rock, Stephen M |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Haomiao Huang. |
|---|---|
| Note | Submitted to the Department of Aeronautics and Astronautics. |
| Thesis | Thesis (Ph.D.)--Stanford University, 2012. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2012 by Haomiao Huang
- 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...