DISSERTATION Sensor Fusion in Time - Triggered Systems

Sensor fusion is the combining of sensory data or data derived from sensory data in order to produce enhanced data in form of an internal representation of the process environment. The achievements of sensor fusion are robustness, extended spatial and temporal coverage, increased confidence, reduced ambiguity and uncertainty, and improved resolution. This thesis examines the application of sensor fusion for real-time applications. The time-triggered approach provides a well suited basis for building real-time systems due to its highly deterministic behavior. The integration of sensor fusion applications in a time-triggered framework supports resource-efficient dependable real-time systems. We present a time-triggered approach for real-time sensor fusion applications that partitions the system into three levels: First, a transducer level contains the sensors and the actuators. Second, a fusion/dissemination level gathers measurements , performs sensor fusion and distributes control information to the actuators. Third, a control level contains a control program making control decisions based on environmental information provided by the fusion level. Using this architecture, complex applications can be decomposed into smaller manageable subsystems. Furthermore, this thesis evaluates different approaches for achieving dependability. These approaches attack the problem at different levels. At the transducer level, we introduce a filter algorithm that performs successive measurements in order to improve the data quality of a single sensor. A different approach improves data by combining data of multiple sensors at the fusion/dissemination level. We propose two sensor fusion algorithms to accomplish this task, the systematic confidence-weighted averaging algorithm and the application-specific robust certainty grid algorithm. The proposed methods are evaluated and used in a case study featuring an autonomous mobile robot. Contents 1 Introduction 1 1.