Abstract
Visual-guided locomotion for snake-like robots is a challenging task, since it involves not only the complex body undulation with many joints, but also a joint pipeline that connects the vision and the locomotion. Meanwhile, it is usually difficult to jointly coordinate these two separate sub-tasks as this requires time-consuming and trial-and-error tuning. In this paper, we introduce a novel approach for solving target tracking tasks for a snake-like robot as a whole using a model-free reinforcement learning (RL) algorithm. This RL-based controller directly maps the visual observations to the joint positions of the snake-like robot in an end-to-end fashion instead of dividing the process into a series of sub-tasks. With a novel customized reward function, our RL controller is trained in a dynamically changing track scenario. The controller is evaluated in four different tracking scenarios and the results show excellent adaptive locomotion ability to the unpredictable behavior of the target. Meanwhile, the results also prove that the RL-based controller outperforms the traditional model-based controller in terms of tracking accuracy.
Highlights
Inspired by real snakes, snake-like robots are designed as a class of hyper-redundant mechanisms in order to achieve the agility and adaptability of their biological counterparts
As our work is related to the perception-driven locomotion of snake-like robots and perception-driven algorithms based on reinforcement learning, we briefly review the state-of-the-art research on both aspects in the following
As a principle approach to temporal decision-making problems, reinforcement learning (RL)-based approaches have been used for solving visual object tracking tasks that aim at finding the target position in contiguous frames and whereby steering the locomotion of an mobile agent
Summary
Snake-like robots are designed as a class of hyper-redundant mechanisms in order to achieve the agility and adaptability of their biological counterparts. Strategies based on Reinforcement Learning (RL) are promising solutions for performing target tracking for a snakelike robot This is because a RL-trained controller can take the visual image directly as the input, while simultaneously fully exploring the locomotion capabilities compared with model-based methods. When traditional methods are used on mobile platforms, target tracking is usually divided into tracking and control sub-tasks, which makes it difficult to tune the pipeline jointly, especially considering the aforementioned motion barrier for snake-like robots To cope with this hard-to-predict tracking and movement complexity, the RL-based control strategies need to map the visual inputs to the joint space directly, in order to perform the corresponding motions, and must operate with adequately defined reward functions to train a policy successfully. We demonstrate that the learned locomotion outperforms the model-based locomotion in terms of tracking accuracy
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
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
