This study introduces a vision-language navigation model for indoor robots. It interprets user prompts and anchors multi-modal context using reinforced alignment, enabling voice-guided autonomous indoor navigation in complex environments.

Olivia Collins, Ethan Reeves, Lily Gardner, Andrew Wells, Chloe Sanders

This paper presents an object re-identification system tailored for dynamic crowds. By combining temporal motion cues with spatial attention, it maintains accurate ID tracking across frames in dense urban surveillance footage.

Chen Hao Lin, Liu Fang Ze, Xu Ming Tao, Zhang Tian Wei, Gao Bo Jian

This study presents a dual-encoder model for depth estimation under extreme lighting. It uses shadow-aware fusion and contrast normalization, maintaining accurate perception in environments with glare, darkness, or reflective surfaces.

Isla Patterson, James Burke, Victoria Willis, Ethan McDowell, Abigail Freeman

This paper proposes a forecasting system for urban crowd management. It integrates video-based motion forecasting with graph neural aggregation, enabling crowd behavior prediction for safety and resource allocation in smart city systems.

Chen Rui Xiang, Liu Bo Xing, Xu Hao Jin, Zhang Ming Cheng, Gao Wen Fang

This work introduces an object detector that quantifies uncertainty in predictions. It uses Bayesian CNNs and Monte Carlo sampling to improve robustness in ambiguous road environments, supporting decision confidence in self-driving cars.

Avery Wells, Mason Doyle, Naomi Spencer, Oliver Harris, Lauren Chapman

This study develops a compact segmentation model for distinguishing vehicle types. Using resolution-aware downsampling and spatial context injection, it balances accuracy and latency in real-time systems for traffic monitoring and self-driving analytics.

Chen Fang Yu, Liu Tian Qiang, Zhang Hao Lin, Xu Jian Cheng, Gao Zhi Liang

This study presents a semantic mapping system for indoor warehouse robots. Dual visual transformers are used to capture object boundaries while memory consolidation enhances persistent labeling for dynamic path optimization in automated inventory handling and navigation workflows.

Chen Rui Shan, Liu Zhen Fang, Xu Long Wei, Zhang Min Hao, Gao Yong Qiang