Influence of IMU parameters on navigation performance in network-based collaborative navigation scenarios

Abstract

Listen

Translate article

The widely used Error State Kalman Filter (ErKF) is known to deliver a suitable navigation solution for most navigation applications. A commonly used architecture is an Inertial Measurement Unit (IMU) that delivers system input to a strapdown algorithm while aiding sensors like Global Navigation Satellite Systems (GNSS) and Barometric Altitude Sensors are used to update the error state. Following the evolution from single-sensor to multi-sensor integration leads to the concept of collaborative sensors to improve the navigation capability of a group of users. The idea is that group members collaboratively share information about their state estimates and sensor data which enables the group to determine an optimal navigation solution for all users. In real-time collaborative navigation system, there are two challenges that need to be solved and they cannot be sharply separated. First, a dynamic sensor network must be established and maintained that, in addition to communication, allows precise time synchronization and at least measurement of the range between nodes (inter-nodal ranging). Second, the estimation of the distributed sensors must be solved, which can be underdetermined and highly nonlinear. To achieve the optimal solution one would have to process the data of all users and additional inter-nodal information (e.g. ranges or direction) in one centralized ErKF. This would however impose significant communication costs on all users and computation effort on the master node that conducts the calculations. A cheaper approach in terms of communication and calculation efforts is a decentralized filter algorithm where only certain information are available to each member of the network. The filter calculations are not conducted by one master node but each member uses their available information to enhance their own position estimate. The navigation solution in this approach will not be as accurate as the centralized solution but more efficient and more robust to disturbances in communication. This paper shows simulation results of a network-based collaborative navigation scenario featuring a decentralized approach for the state estimation. Additional to his own state vector, a network member is assumed to have knowledge about the inter-nodal range of itself to a changing number of other members of the same network as well as their position estimates. Certain nodes have independent knowledge about their absolute position, i.e. are considered to have GNSS information available. On the basis of these information, each member can enhance their position estimate. The behavior of the network in cases of GNSS and/or inter-nodal range outages are shown with a special focus on the influence of IMU performance parameters. For demonstration several IMU performance classes are simulated and results compared. As a benchmark, we compare our results with results generated by the above mentioned centralized Kalman filter.