Mock-measurement based performance evaluation of inertial navigation without knowing ground truth

As an important navigation technology, the inertial navigation system (INS) has been widely used in various applications and many INS algorithms have been proposed. Their performance evaluation is crucial for evaluating and improving the algorithms, where the ground truth is assumed known. However, knowledge of the ground truth is hard to acquire, and this presents a challenge to performance evaluation. In this paper, a truth-knowledge free approach to performance evaluation of INS algorithms is proposed. In this approach, we generate some mock measurements from the INS outputs and judging the performance by checking how close the mock measurements are to the real measurements. The INS algorithm whose mock measurements are closer to the real measurements is preferred. Simulations and related analysis are provided to illustrate and validate our proposed evaluation approach.