Boosted Model Tree-Based Behavioral Modeling for Digital Predistortion of RF Power Amplifiers

Abstract

In this article, we propose a new behavioral modeling approach, called boosted model tree, to characterize and compensate for the complex nonlinear distortions induced by wideband high-efficiency radio frequency power amplifiers. With the proposed model, the input data are classified into different zones by decision trees and each zone is assigned separate submodels. We also employ a model boosting technique to build multiple parallel tree structures that jointly model the desired nonlinear behavior. By designing dedicated optimization procedures, both tree structures and submodel coefficients can be efficiently identified. It is demonstrated that the combination of piecewise and parallel structures provides a powerful and hardware-efficient way to model nonlinear memory effect and cross terms. Based on the experimental results, the proposed method can achieve improved linearization performance with low hardware complexity under challenging wideband predistortion scenarios.

Digital Predistortion of RF Power Amplifiers With Phase-Gated Recurrent Neural Networks

“…For example, the GMP model [4] adopts polynomial terms in the form of |x n−m−l | p x n−m to characterize PA behavior. Yet, the PA memory effect in wideband applications will inevitably involve a large number of memory samples, but the polynomial operator may be incapable to learn very complex nonlinear functions [24].…”

Particle swarm optimization‐XGBoost‐based modeling of radio‐frequency power amplifier under different temperatures

“…However, although ANN has shown good nonlinear fitting ability, it is easy to overfit as it adopts the empirical risk minimization (ERM) principle 12 . In addition, the training of ANN requires a large amount of data support, and it is difficult for actual RF PA to provide such a large amount of data, which will also limit the application of ANN in RF PA 13 …”

“…However, the modeling complexity of SVR is relatively high, requiring a significant number of computational resources and time 12 . Li et al established a boosting tree model for PA by parallelizing multiple decision trees and built a higher‐precision PA model with fewer resources 13 . However, decision trees are prone to overfitting, and simply parallelizing decision trees exacerbates this advantage 14 .…”

“…12 Li et al established a boosting tree model for PA by parallelizing multiple decision trees and built a higher-precision PA model with fewer resources. 13 However, decision trees are prone to overfitting, and simply parallelizing decision trees exacerbates this advantage. 14 Dikmense compared the effects of multiple machine learning models such as gradient boosting, SVR, decision trees, random forest, and ANN on PA modeling and found that gradient boosting is optimal in modeling accuracy and speed.…”

Machine Learning-Aided Piece-Wise Modeling Technique of Power Amplifier for Digital Predistortion

“…a single model for the entire range of output power because of varying behavior at different power levels. Thus, piecewise (PW) polynomial-based models have been shown to be quite effective in modeling and linearizing PAs with strong nonlinear effects [12][13][14][15]. In [12], a vector switched model has been proposed where the input data samples are classified by a computationally expensive k-means clustering algorithm based on their envelope.…”