The article discusses the development of DogSurf, a quadruped robot that assists visually impaired individuals in navigating their environment. DogSurf is able to detect slippery surfaces and uses audio and haptic feedback to alert the user when to stop moving. The robot is powered by a GRU-based neural network, which has a mean accuracy of 99.925% for multiclass surface classification. The data and code used in the development of DogSurf have been made publicly available.
Publication date: 5 Feb 2024
Project Page: https://doi.org/XX.XXXX/XXXXXXX.XXXXXXX
Paper: https://arxiv.org/pdf/2402.03156
