A student team named IN²Bot from the School of Automation at Beijing Institute of Technology (BIT) has won championships at two leading international embodied AI competitions.

They took first place at the ManiSkill-ViTac 2026 Real-World Language-Guided Bimanual Vision-Tactile Manipulation Challenge as well as at the mobile robot autonomous navigation challenge of the IEEE International Conference on Robotics and Automation.

The ManiSkill-ViTac challenge of the CVPR, the largest and most influential conference in computer vision, focuses on fine laboratory operations, setting long-horizon bimanual tasks such as cleaning test tubes and grinding materials, in which robots must complete precise manipulations guided largely by touch under limited visual input.

The BIT team proposed a visuotactile fusion and action-memory scheme, developing its own tactile encoder that enables robots to build a "motion-touch" perception capability, and introducing an action memory module to overcome forgetting in long-sequence tasks. The team won by a margin of 700 points.

The other competition, the mobile robot autonomous navigation challenge, tests robots under three extreme conditions: satellite denial, absence of prior maps, and extremely narrow unknown spaces, simulating real-world scenarios such as underground search and rescue and rubble traversal.

The BIT team approached the problem on three levels: cognition-inspired perception, intelligent decision-making and fine-grained control, building a traversability-aware spatial reasoning field, proposing a local reactive planning method and introducing a safety-barrier trajectory control strategy, which allowed its robot to thread through tight spaces smoothly. The team recorded the highest overall score in both the simulation evaluation and the on-site final.

Going forward, BIT will continue to integrate frontier research with talent cultivation and contribute to the development of embodied intelligent robotics.