Read the full-text of this article for free

Title: Obstacle-free robot path planning based on variational autoencoder and generative networks

Authors: Laide Guan; Bole Li

Addresses: School of Mechanical and Electrical Engineering, Liuzhou Vocational and Technical College, Liuzhou 545006, China ' School of Mechanical and Electrical Engineering, Liuzhou Vocational and Technical College, Liuzhou 545006, China

Abstract: Robot route planning becomes crucial in intelligent navigation systems given the fast advancement of automation technologies. Concerning efficiency and resilience in handling challenging dynamic settings, traditional path-planning approaches have several restrictions. Based on variational autoencoder (VAE) and generative adversarial network (GAN), this work presents a path planning model, VAE-GAN PathNet, to handle this challenge. By combining the benefits of VAE in latent space modelling and the capacity of GAN in path optimisation, the model essentially increases the quality, smoothness and obstacle avoidance performance of path planning. This work uses the Stanford Drone Dataset and the ROS Path Planning Dataset to validate the efficacy of the model using trials. In terms of path length, obstacle avoidance performance, path smoothness and computing time, VAE-GAN PathNet beats conventional path planning algorithms experimental data demonstrate.

Keywords: variational autoencoder; VAE; generative adversarial network; GAN; path planning; obstacle avoidance; path smoothness.

DOI: 10.1504/IJICT.2025.145704

International Journal of Information and Communication Technology, 2025 Vol.26 No.7, pp.17 - 31

Received: 10 Feb 2025
Accepted: 19 Feb 2025
Published online: 15 Apr 2025 *