Reinforcement Learning Control of a Forestry Crane Manipulator

Andersson, Jennifer; Bodin, Kenneth; Lindmark, Daniel; Servin, Martin; Wallin, Erik

doi:10.48550/arxiv.2103.02315

Cited by 2 publications

(3 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another set of applications of RL for mobile machines are found in [100] where it is used, based on camera, lidar, and motion and force sensors, to perform the bucket loading in underground mine applications in a multi-objective target, including the maximisation of bucket loading; in [101], it is trained to control the motion of a forestry crane while considering the minimisation for energy consumption; in [102], it is used for the trajectory tracking control of the motion of an excavator arm aiming for autonomous application, where the controller generates the valve control signals directly; and in [103], RL is used to adapt for the real world conditions of a network trained to control the motion of the actuators of a wheel loader during bucket filling.…”

Section: Reinforcement Learningmentioning

confidence: 99%

On Machine Learning-Based Control for Energy Management in Construction Machines

Raduenz

2022

Linköping Studies in Science and Technology. Dissertations

View full text Add to dashboard Cite

High energy efficiency is a key requirement for modern construction machinery. This is because of stricter environmental targets, electrification, and reduction of operation costs. To meet this requirement, the powertrain architectures of the machines are becoming increasingly complex, for example through hybridisation of drivetrain and work functions, or with improved hydraulic systems. However, the more complex the architecture is, the harder the management of splitting power between different sources and consumers. The number of work functions, operating environments, and tasks these machines engage in, along with the added degrees of freedom with respect to how energy can be recovered, exchanged, and reused, makes them unique. Therefore, the development of control strategies for energy management in such machines requires specific research and development with their architecture and application in focus. This doctoral thesis presents an analysis of two methods for the development of machine learning-based energy management strategies for construction machines. One is based on supervised learning and the other on reinforcement learning. The methods use optimisation to find optimised solutions for the control problem of the systems and machine learning for learning and implementing the control decisions. In both methods, models of the physical systems are used for the learning and training. The thesis highlights and confirms, with experimental results, the potential of such methods to derive control strategies for these machines. The studied methods can learn and implement improved control decisions in the real systems that result in the potential for increased efficiency. At the same time, their robustness is shown in the application to unseen scenarios during training, although that does not eliminate the need for further training in the real systems after deployment. The thesis also increases the comprehensiveness on energy management for construction machines. The thesis was completed in a double-degree format between the

show abstract

Section: Reinforcement Learningmentioning

confidence: 99%

On Machine Learning-Based Control for Energy Management in Construction Machines

Raduenz

2022

Linköping Studies in Science and Technology. Dissertations

View full text Add to dashboard Cite

show abstract

“…This technique is unsuitable for log piles, since the masked pile can be blob-shaped, making it impossible to distinguish individual instances geometrically. More related to forwarders and wood loaders, motion planning to grasp logs is analyzed in [8]. There, they used reinforcement learning to control the crane motion of a forwarder.…”

Section: A Instance Segmentation Architecturesmentioning

confidence: 99%

“…Taking advantage of these technological advances, unmanned logging machines are getting closer to reality. However, previous work on autonomous log handling [8] assume that the position and orientation of the logs are known, which is generally not the case.…”

Section: Introductionmentioning

confidence: 99%

Instance Segmentation for Autonomous Log Grasping in Forestry Operations

Fortin¹,

Gamache²,

Grondin³

et al. 2022

Preprint

View full text Add to dashboard Cite

Wood logs picking is a challenging task to automate. Indeed, logs usually come in cluttered configurations, randomly orientated and overlapping. Recent work on log picking automation usually assume that the logs' pose is known, with little consideration given to the actual perception problem. In this paper, we squarely address the latter, using a datadriven approach. First, we introduce a novel dataset, named TimberSeg 1.0, that is densely annotated, i.e., that includes both bounding boxes and pixel-level mask annotations for logs. This dataset comprises 220 images with 2500 individually segmented logs. Using our dataset, we then compare three neural network architectures on the task of individual logs detection and segmentation; two region-based methods and one attention-based method. Unsurprisingly, our results show that axis-aligned proposals, failing to take into account the directional nature of logs, underperform with 19.03 mAP. A rotation-aware proposal method significantly improve results to 31.83 mAP. More interestingly, a Transformer-based approach, without any inductive bias on rotations, outperformed the two others, achieving a mAP of 57.53 on our dataset. Our use case demonstrates the limitations of region-based approaches for cluttered, elongated objects. It also highlights the potential of attention-based methods on this specific task, as they work directly at the pixel-level. These encouraging results indicate that such a perception system could be used to assist the operators on the short-term, or to fully automate log picking operations in the future.

show abstract

Reinforcement Learning Control of a Forestry Crane Manipulator

Cited by 2 publications

References 14 publications

On Machine Learning-Based Control for Energy Management in Construction Machines

On Machine Learning-Based Control for Energy Management in Construction Machines

Instance Segmentation for Autonomous Log Grasping in Forestry Operations

Contact Info

Product

Resources

About