| Autonomous robotic operation in an unstructured or partially known environment requires sensing and sensor-based control. To overcome the problems with current "eye-in-hand" systems, miniature amplitude-based, infra-red proximity sensors are being studied. Obtaining position and velocity estimates of a rigid body with these sensors is a non-linear parameter and state estimation problem. Among the methods examined in simulation, Extended Kalman Filtering (EKF) was selected for implementation. A novel approach for object localization was developed in which the object geometry is known, sensing is performed by a proximity sensing network (PSN) and the object's unknown reflective properties are estimated on-line. The method has been tested extensively in simulation and experiments in which a target object's planar position and velocity were successfully estimated. To the author's knowledge this is the first time amplitude based infra-red sensors have been used to estimate a rigid body's unknown trajectory. |
