Adaptive vision-based guidance law with guaranteed performance bounds for tracking a ground target with time-varying velocity

This work extends the earlier results of authors on vision-based tracking of a ground vehicle moving with unknown time-varying velocity. The follower UAV is equipped with a single camera. The control objective is to regulate the 2D horizontal range between the UAV and the target to a constant. The extension in this paper has two distinct features. The earlier developed guidance law used the estimates of the target's velocity obtained from a fast estimation scheme. In this paper, we prove guaranteed performance bounds for the fast estimation scheme and explicitly derive the tracking performance bound as a function of the estimation error. The performance bounds imply that the signals of the closed-loop adaptive system remain close to the corresponding signals of a bounded closed-loop reference system both in transient and steady-state. The reference system is introduced solely for the purpose of analysis. This paper also analyzes the stability and the performance degradation of the closed-loop adaptive system in the presence of out-of-frame events, when continuous extraction of the target's information is not feasible due to failures in the image processing module. The feedback loop is then closed using the frozen estimates. The out-of-frame events are modelled as brief instabilities. A sufficient condition for the switching signal is derived that guarantees graceful degradation of performance during target loss. The results build upon the earlier developed fast estimation scheme of the target's velocity, the inverse-kinematics-based guidance law and insights from switching systems theory.