Research on hybrid estimation algorithm for dynamic center-of-mass parameters of tracked agricultural transport platform
- 1College of Mechanical and Electrical Engineering, Changchun University of Science and Technology
- 2College of Instrumentation and Electrical Engineering, Jilin University
- 3College of Artificial Intelligence, Changchun University of Science and Technology
Abstract
To address dynamic centroid shift and reduced control accuracy caused by complex terrain and multi-device collaborative loading on tracked agricultural transport platforms, a hybrid estimation algorithm (CBA-IPSO-UKF) is proposed. This algorithm is designed for platforms operating in unstructured outdoor environments and combines a deep learning-based model with intelligent optimization techniques. First, the limitations of traditional UKF algorithms—which depend heavily on physical models and struggle with strong nonlinear disturbances—are overcome by developing a CNN-BiLSTM hybrid model with attention mechanisms. This model replaces conventional state and observation equations. Second, to reduce systematic errors from model mismatch, a constant-coefficient compensation vector is incorporated into the estimation framework. An improved particle swarm optimization algorithm, enhanced with genetic algorithm crossover and mutation operators, is used to optimize global parameters and minimize residuals. Real-vehicle data were collected from a self-developed tracked agricultural platform, resulting in a dataset of 18, 000 valid samples. Comparative experiments were carried out under typical circular driving conditions in unstructured environments. The results demonstrate that the proposed algorithm achieves high accuracy and steady-state performance in both attitude and velocity estimation. Specifically, the mean absolute errors for roll angle, pitch angle, and heading angle were 0.0917°, 0.1352°, and 3.1458°, respectively. The MAE for the three-axis velocities reached 0.0622 m/s, 0.0616 m/s, and 0.0209 m/s, respectively. These results show significant improvements over traditional algorithms like UKF and ACKF. This study offers both theoretical support and practical engineering solutions for real-time state perception and the stability control of autonomous operations in complex environments for intelligent agricultural machinery.
Related
- The performance of directional area scattering factor proxies for correcting canopy structural and soil effects on leaf chlorophyll content estimation varies with canopy conditions in agricultural systems
- Estimation of agricultural carbon abatement costs using a marginal-productivity-based production-input model: evidence from Fujian Province, China
- A Matrix-Based Sensor Fusion Algorithm for Nine-D.o.F. IMU Attitude Estimation
- GCOM-C/SGLI ground-area-based chlorophyll content estimation: a computationally efficient algorithm free from LAI and vegetation classification
- Highlighting the strengths of new wearable fNIRS, precision accuracy microgenetic research designs, and new NeuroDOT processing algorithm to measure change over time in clinical populations for Speech-Language and Hearing Researchers
BibTeX
@article{wang2026research,
title = {{Research on hybrid estimation algorithm for dynamic center-of-mass parameters of tracked agricultural transport platform}},
author = {Yitian Wang and Daohong Chong and Jinxin Luo and Longkai Liang and Yangyang Zou},
journal = {Frontiers in Plant Science},
year = {2026},
month = oct,
doi = {10.3389/fpls.2026.1928457},
url = {https://www.frontiersin.org/articles/10.3389/fpls.2026.1928457/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Research on hybrid estimation algorithm for dynamic center-of-mass parameters of tracked agricultural transport platform
AU - Yitian Wang
AU - Daohong Chong
AU - Jinxin Luo
AU - Longkai Liang
AU - Yangyang Zou
PY - 2026
DA - 2026/10/13
JO - Frontiers in Plant Science
DO - 10.3389/fpls.2026.1928457
UR - https://www.frontiersin.org/articles/10.3389/fpls.2026.1928457/full
AB - To address dynamic centroid shift and reduced control accuracy caused by complex terrain and multi-device collaborative loading on tracked agricultural transport platforms, a hybrid estimation algorithm (CBA-IPSO-UKF) is proposed. This algorithm is designed for platforms operating in unstructured outdoor environments and combines a deep learning-based model with intelligent optimization techniques. First, the limitations of traditional UKF algorithms—which depend heavily on physical models and struggle with strong nonlinear disturbances—are overcome by developing a CNN-BiLSTM hybrid model with attention mechanisms. This model replaces conventional state and observation equations. Second, to reduce systematic errors from model mismatch, a constant-coefficient compensation vector is incorporated into the estimation framework. An improved particle swarm optimization algorithm, enhanced with genetic algorithm crossover and mutation operators, is used to optimize global parameters and minimize residuals. Real-vehicle data were collected from a self-developed tracked agricultural platform, resulting in a dataset of 18, 000 valid samples. Comparative experiments were carried out under typical circular driving conditions in unstructured environments. The results demonstrate that the proposed algorithm achieves high accuracy and steady-state performance in both attitude and velocity estimation. Specifically, the mean absolute errors for roll angle, pitch angle, and heading angle were 0.0917°, 0.1352°, and 3.1458°, respectively. The MAE for the three-axis velocities reached 0.0622 m/s, 0.0616 m/s, and 0.0209 m/s, respectively. These results show significant improvements over traditional algorithms like UKF and ACKF. This study offers both theoretical support and practical engineering solutions for real-time state perception and the stability control of autonomous operations in complex environments for intelligent agricultural machinery.
SN - 1664-462X
ER -