Abstract
This paper describes a method of estimating the traversability of plant parts covering a path and navigating through them for mobile robots operating in plant-rich environments. Conventional mobile robots rely on scene recognition methods that consider only the geometric information of the environment. Those methods, therefore, cannot recognize paths as traversable when they are covered by flexible plants. In this paper, we present a novel framework of image-based scene recognition to realize navigation in such plant-rich environments. Our recognition model exploits a semantic segmentation branch for general object classification and a traversability estimation branch for estimating pixel-wise traversability. The semantic segmentation branch is trained using an unsupervised domain adaptation method and the traversability estimation branch is trained with label images generated from the robot's traversal experience during the data acquisition phase, coined traversability masks. The training procedure of the entire model is, therefore, free from manual annotation. In our experiment, we show that the proposed recognition framework is capable of distinguishing traversable plants more accurately than a conventional semantic segmentation with traversable plant and non-traversable plant classes, and an existing image-based traversability estimation method. We also conducted a real-world experiment and confirmed that the robot with the proposed recognition method successfully navigated in plant-rich environments.
Highlights
The mobile robot technologies have come nearly to the level of practical realization and commercialization in some situations
To take full advantage of the incomplete information of the traversability masks, we present a deep neural network (DNN) architecture and its training method inspired by PU learning [11], where a model is trained with labeled positive examples and unlabeled examples including both positive and negative ones
For the training of Traversability Estimation Module (TEM), we introduce a training method inspired by PU learning [11] in order to take full advantage of the information given by the traversability masks
Summary
The mobile robot technologies have come nearly to the level of practical realization and commercialization in some situations. Self-driving cars are already being tested in public areas [1, 2]. Another example is the service robots operating in public places such as airports and stores [3, 4]. Those applications mainly target structured environments such as urban areas and the inside of buildings. In plantrich environments such as agricultural fields and forest paths, it is more difficult for robots to autonomously navigate than in structured environments because the paths are possibly covered by plant parts such as branches and leaves, which, though, can be driven through by the robots. Most of conventional mobile robots only consider the presence of objects and do not consider such traversable objects
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call