Adaptive Excavation Automation in Complex Soil Environments Using Reinforcement Learning

This paper presents an advanced excavation automation framework that improves trajectory efficiency and improves operational stability in diverse and complex soil conditions. The framework integrates the fundamental equation of earthmoving (FEE) with the proximal policy optimization (PPO) algorithm to enable adaptive trajectory planning during excavation. Using these estimates, the PPO algorithm iteratively optimizes excavation strategies to minimize applied force and improve trajectory efficiency, contributing to potential energy savings. The validation of the proposed framework is performed through simulations conducted under varying soil conditions and initial height configurations. The results show an 18.1 % reduction in total excavation resistance compared to baseline models, achieving a mean absolute error (MAE) of 0.065 m. These findings confirm the effectiveness of the framework in reducing excavation resistance and improving task precision in simulated environments. Note to Practitioners—This paper is inspired by the lack of applicable excavation mechanisms due to the variability of soil conditions and proposes an excavation framework that combines soil resistance modeling and artificial intelligence (AI) to improve operational efficiency in various soil environments. The system models key soil parameters and dynamically adjusts excavation trajectories to minimize resistance and improve leveling accuracy. The focus is on improving robustness to flexibly adapt to diverse soil conditions. Industry professionals in construction, mining, and infrastructure development can utilize this framework to reduce energy consumption and improve operational precision. Experimental results demonstrate that the proposed solution effectively reduces excavation resistance and improves task accuracy, demonstrating its applicability to real-world automated environments. However, this approach assumes static soil properties during single cycles, which may pose limitations in rapidly changing environments. In addition, obtaining precise soil property data in real-world scenarios remains a challenge. Future research will address the design of excavation algorithms based on real-time sensor data integration for soil property estimation.