Towards Autonomous Wood-Log Grasping with a Forestry Crane: Simulator and Benchmarking

TL;DR

This work tackles autonomous manipulation with large-scale forestry cranes by developing an open-source MuJoCo-based simulator that models an 8-DoF crane and varied-diameter logs. It introduces a latent-MDP framework to handle log diameter variability and a modified PPO (mPPO) that uses a Beta action distribution for safe, bounded control, trained across many parallel environments. The results show high success rates (>96%) across log sizes and randomized configurations, along with insights into the learned grasping strategy and sim-to-real considerations. The paper provides a practical benchmark and analyzes sim-to-real transfer, highlighting the benefits and remaining challenges for real-world deployment of autonomous forestry manipulation systems.

Abstract

Forestry machines operated in forest production environments face challenges when performing manipulation tasks, especially regarding the complicated dynamics of underactuated crane systems and the heavy weight of logs to be grasped. This study investigates the feasibility of using reinforcement learning for forestry crane manipulators in grasping and lifting heavy wood logs autonomously. We first build a simulator using Mujoco physics engine to create realistic scenarios, including modeling a forestry crane with 8 degrees of freedom from CAD data and wood logs of different sizes. We further implement a velocity controller for autonomous log grasping with deep reinforcement learning using a curriculum strategy. Utilizing our new simulator, the proposed control strategy exhibits a success rate of 96% when grasping logs of different diameters and under random initial configurations of the forestry crane. In addition, reward functions and reinforcement learning baselines are implemented to provide an open-source benchmark for the community in large-scale manipulation tasks. A video with several demonstrations can be seen at https://www.acin.tuwien.ac.at/en/d18a/

Figures (6)

• Figure 1: (a) Snapshot of the simulated environment. While the logs' length is fixed to $l=2.75m$, their diameters are varied in the range $d=[0.3,\:0.8]$ m, (b) Real crane setup in outdoor environment.
• Figure 2: Schematic of the forestry crane ecker2022iterative.
• Figure 3: Details of variables used for constructing the observations and reward function.
• Figure 4: Overview of the learning process. $m$ randomized environments with different wood log sizes and poses are generated by our crane simulator, presented in Subsection \ref{['sec: b simulator']}.
• Figure 5: Evolution of cumulative rewards over $12000$ update steps of the optimizer.