3 papers indexed
Yangyang Ning, Shu Liang, Quanbo Ge, Tianchen Deng, Yuhua Qi, Shenghai Yuan
Learning is increasingly introduced into visual-inertial odometry (VIO), ranging from learned feature front-ends to learning-dominant motion and geometry estimation. However, learning more of the pipeline does not necessarily improve robustness when deployment conditions differ f…
Yuwen Liao, Yihang Lan, Yizhuo Yang, Ruimeng Liu, Xinhang Xu, Shenghai Yuan, et al.
Social navigation requires the robot to reason and respond in complex real-world environments. While recent works attempt to incorporate human-level intelligence into robot planning using large Vision-Language Models (VLMs), end-to-end frameworks often create an unpredictable bla…
Huzhenyu Zhang, Shenghai Yuan, Wenrui Yan, Li Ma, Hengjie Li, Jingcheng Pang, et al.
Data-driven embodied AI is rapidly transitioning into a paradigm that scales training through massively parallel simulation, where GPU-accelerated simulators serve as the foundational data infrastructure. However, as computational throughput scales, the underlying trade-offs betw…