Theory for a mobile robot intercepting projectiles above or below the horizon.

摘要

When a fielder runs to catch a ball, a driver navigates to avoid a collision, or a pedestrian intercepts a fellow walker, they each utilize unconscious spatial orienting principles and other automatic perceptual control mechanisms. In this dissertation, the study of human interception of stationary or moving targets is applied to mobile robot navigation. Using perceptual information, geometric principles are derived to plan an interceptive path. These new control algorithms provide a simple, real-time method to guide a mobile robot to intercept either a stationary or a moving target.; When catching a moving target like a projectile that approaches from either above or below the horizon, humans are surprisingly accurate at intercepting the object. This research is the first to utilize high-precision, marker-based, motion capture technology to unambiguously measure head, body, and target position during high-speed interception tasks. Data analysis suggests that humans use two strategies to control their motion within a plane. One is the optical acceleration cancellation (OAC) strategy, which reduces the gap between the pursuer and the target on the horizontal plane using the tangent or cotangent function. The other strategy is the lateral alignment strategy, which maintains the gaze direction projected on the horizontal plane to be constant. The OAC model can determine a set of possible pursuer positions within the horizontal plane at every time instant. Together with the lateral alignment strategy, a unique position can be determined.; Active and passive visual servo controllers are designed and simulated. The active controller is superior in the fly ball interception task. The passive controller is better in the ground target interception task. An alternative control model for ground target interception is also proposed which directly uses a visual error to control the robot motion. These new interception models are implemented on a high-speed mobile robot using an automatic real-time control system. In experimentation, the robot can intercept ground targets and simple fly ball trajectories.; This dissertation describes and validates an interception model based on human perceptual principles. It not only provides insights into human interceptive behaviors, but also enhances the path-planning ability for mobile robots.