Active Target Tracking Using Bearing-only Measurements With Gaussian Process Learning
This paper studies the tracking problem of a target with the partially unknown motion model by an active agent with bearing-only measurements using Gaussian process learning. To address this problem, a learning-planning-control framework is proposed. First, to learn and predict the target motion und...
Gespeichert in:
Hauptverfasser: | , , , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | This paper studies the tracking problem of a target with the partially
unknown motion model by an active agent with bearing-only measurements using
Gaussian process learning. To address this problem, a learning-planning-control
framework is proposed. First, to learn and predict the target motion under mild
assumptions, a Gaussian-process-based scheme is proposed, and a probabilistic
uniform prediction error bound can be rigorously proved. Second, by analyzing
the data dependence of the posterior covariance, we obtain an optimal relative
trajectory to achieve efficient sampling. Third, to realize efficient learning,
a controller to track the planned path is proposed based on the learned target
motion, which can provide guaranteed tracking performance. Theoretical analysis
is conducted to prove the the given probabilistic error bounds. Numerical
examples and comparison with other typical methods verify the feasibility and
superior performance of our proposed framework. |
---|---|
DOI: | 10.48550/arxiv.2410.18669 |