Adaptive and reconfigurable data fusion architectures in positioning navigation systems.

摘要

In automotive positioning systems, at any time, any of the sensors can break down or stop sending information, temporarily or permanently. As a result, this may lead to situations with safety, health, financial or legal implications. Although, good design practice tends to minimize the occurrence of sensor faults and failures, it is recognized that such events do occur. In such case, faulty or failed sensor must be detected and isolated so that the faulty data will not corrupt the global estimates, and the system must finally be able to reconfigure itself so as to overcome the deficiency caused by the fault. In brief, a navigation system must be robust and adaptive.;Experiments show that the proposed solutions can compensate for most of the errors associated with sensors faults or performance degradations, and that the resulting positioning accuracy is improved significantly.;Keywords. Positioning, Vehicle Navigation, Sensor Fusion, GPS, Kalman Filter, Fault Detection, Robustness;In this thesis, several sensor fault adaptive data fusion architectures are proposed to deal with the above case. These approaches apply Kalman filters in combination with fault detection so as to produce robust positioning modules. These modules should be capable of handing situation where GPS input is corrupted or unavailable, one or more of others position sensors are faulty or paralytic. The working principle is to modify the gains of the Kalman filter based on the normalized errors between the estimate states and observations. To test the proposed architecture, various sensors faults or performance degradations are implemented and simulated.