| dc.description.abstract | Symmetric turning control is the typical design choice for most machines. However, historical examples of asymmetric machine design, as well as examples of asymmetry in nature, suggest that asymmetric turning may be a potential advantage in adversarial applications. For instance, aircraft of World Wars I and II were plagued by asymmetric turning controls as a result of gyroscopic forces from the rotating engine. Pilots of the time actually believed this to be a feature, not a bug, suggesting that the asymmetric turning improved strategic evasion and pursuit during battle. As autonomous robots become increasingly critical in military operations, it is imperative that we endow them with strategic designs for better performance. We seek to understand if asymmetric turning is an advantageous design.
Using Karaman and Frazzoli's sample-based algorithm for pursuit-evasion games, software simulates robot motion planning in an asymmetric Dubins state space to observe how asymmetric turning influences agent success. We demonstrate mathematically that the Dubins interval path solutions are applicable to asymmetric Dubins vehicles, as both are utilized within the simulation. The Open Motion Planning Library (OMPL) is leveraged to implement the pursuit-evasion game algorithm. To simulate asymmetric action, agents are assigned varying degrees of asymmetric turning constraints, such that as one turn sharpens, the other broadens. Agents then compete in a pursuit-evasion game. Pursuit-evasion games are simulated across a range of asymmetric turning match-ups and agent starting positions.
Results show that pursuer success increases as its asymmetry increases. Evader success remains constant, regardless of asymmetric turning influence. Furthermore, the advantages of asymmetric turning can be further augmented when considered in conjunction with relative agent starting position. The results of this research inform more intelligent machine design strategies for vehicles in dynamic spaces. | en |
