Gaussian process-based visual pursuit control with unknown target motion learning in three dimensions

Abstract

In this paper, we propose an observer-based visual pursuit control integrating three-dimensional target motion learning by Gaussian Process Regression (GPR). We consider a situation where a visual sensor equipped rigid body pursuits a target rigid body whose velocity is unknown but dependent on the target's pose. We estimate the pose from visual information and propose a Gaussian Process (GP) model to predict the target velocity from the pose estimate. We analyse stability of the proposed control by showing that estimation and control errors are ultimately bounded with high probability. Finally, simulations illustrate the performance of the proposed control schemes even if the visual measurement is corrupted by noise.

Cooperative Target Tracking by Multiagent Camera Sensor Networks via Gaussian Process

“…From Assumption 2, the Gaussian process can provide the confidence interval of the estimation. In [27], [28], and [42] the probabilistic uncertainty bound with the information gain has been considered. Another approach for evaluating the estimation error is to use nonprobabilistic analysis to obtain a hard error bound.…”

“…The tracking performance in terms of the ultimate boundedness is shown by analyzing the mean and the marginal variance of the Gaussian process. Moreover, the authors in [28] have proposed an observerbased method for visual pursuit control. In [28], the control law to learn the target velocity with the Gaussian process model was considered.…”

“…Moreover, the authors in [28] have proposed an observerbased method for visual pursuit control. In [28], the control law to learn the target velocity with the Gaussian process model was considered. In [29], the authors proposed an iterative learning method for underactuated robotic systems with the Gaussian process regression.…”

Visual Pursuit Control based on Gaussian Processes with Switched Motion Trajectories

“…and the vector form g = [p ξ θ] (Omainska et al, 2021). We consider the scenario of a moving target and cameraequipped mobile robot.…”

“…In our previous work (Omainska et al, 2021;Yamauchi et al, 2021) we assumed that targets move according to their environment. A disadvantage about the proposed solutions is that the GP model can express only one particular motion trajectory.…”