A novel vision-language model-based approach for socially compliant robot navigation in human-centric environments.
A memory-enabled deep reinforcement learning framework is proposed to enhance autonomous robot navigation in diverse and crowded pedestrian environments by leveraging long-term memory, modeling human-robot interactions, and integrating global planning.
A generative adversarial network (GAN) model is proposed to enable mobile robots to generate socially adaptive navigation paths that are similar to human demonstration paths in human-robot interaction environments.
A group-based social navigation framework (GSON) that leverages the visual reasoning capabilities of Large Multimodal Models (LMM) to enable mobile robots to perceive and exploit the social group structure of their surroundings, and generate socially appropriate motions that avoid disrupting the social context.