.png){kind=icon}
.png){kind=icon}
Return
Optimization design method for typical grassland perception robot system
.png){kind=icon}
PRE
AI
DOI：10.1007/s12293-024-00433-3.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This study introduced a perceptive robot system optimization method and framework designed specifically for grassland ecological monitoring, focusing on improving data collection efficiency and prediction accuracy. This study first took the grassland perceptive robot as the object to establish its kinematics and dynamics model. A component-based perceptive robot body mechanism parameter optimization method was proposed on this basis. This method included an anisotropic dynamics optimization algorithm for angles and a global dynamics optimization algorithm. The optimization algorithm designs functions that target the maximum torque of the drive joint and torque fluctuations and quantitatively express the kinematic sensitivity. This study took the particle swarm optimization (PSO) and the multiple objective PSO (MOPSO) as examples to solve the optimization model. The optimization results show that the optimized mechanism parameters can improve the perceptive robot's global dynamics and kinematics performance. This study optimized the AGB (above-ground plant biomass) estimation model by optimizing the perception of the robot body. This study implemented a modular AGB estimation framework to select the most appropriate algorithm based on specific data sets and research needs. This study highlights the importance of surface vegetation cover and plant height as key predictor variables for AGB estimation. Combining multi-modal data can significantly improve the model's prediction accuracy. The AGB estimated R2 for actual grassland trials was 0.78. The system optimization method in this study reduced the time consumption by 70% and 30% respectively compared with the traditional mowing method and the vehicle-mounted sensor method. The above models and algorithms use the componentization and modularization ideas of the robot system. Future research will explore more variables to optimize robot performance and estimation model accuracy further, promoting the development of ecological monitoring technology.
Keywords:
Above ground biomass estimation
Perception robot
Overall dynamic performance
Deep learning
Multi-modal information