This paper presents a material recognition framework using multi-spectral vision. Fused spectral features are processed with graph convolutions, allowing precise classification of similar textures and materials in autonomous industrial inspection settings.

Namit Raghunandan Nair, Savannah Joy McBride, Shen Jia Hao, Kaito Hideaki Fujimoto, Lillian Belle Carpenter

This study proposes a loop closure detection system. Scene memory embedding and perceptual hashing allow robots to identify revisited places with low latency and high robustness under viewpoint and appearance variation.

Dev Arnav Rajput, Felicity Dawn Simmons, Lin Zhi Peng, Daichi Masato Hoshino, Victoria Anne Beckett

This paper introduces a few-shot sorting framework. Using metric-based embeddings and task-driven data augmentation, it enables robots to rapidly learn new categories and perform object sorting with minimal supervision in dynamic, real-world settings.

Siddhant Rishi Agarwal, Amelia Rose Bradford, Zhao Fang Hui, Shuji Hiro Tanaka, Ruby Faith Garrison

This study presents a framework for reconstructing dynamic environments through temporal visual keypoint tracking. Predictive trajectory modeling enhances navigation planning for robots operating in partially occluded or frequently changing spaces with minimal prior mapping data.

Jayant Harinder Kaul, Martha Eloise Chambers, Chen Rong Xi, Thomas Isaac Burnham, Meilin Anzhuo Liu

This paper introduces an efficient 3D detection pipeline combining sparse LiDAR data with visual cues through transformer networks. It significantly improves object localization and classification accuracy for autonomous navigation in both structured and unstructured terrains.

Haruto Kenshi Watanabe, Grace Eleanor Mansfield, Arjun Ragav Deshmukh, Victor Harold Glenn, Li Chun Wei

This study proposes a cross-modal fusion model for object matching under low-light scenarios. Attention mechanisms dynamically prioritize thermal and visible spectrum data, enhancing accuracy in identifying and tracking targets in visually compromised environments.

Sofia Renee Whitmore, Deepanraj Kishore Iyer, Fang Zhou Ping, Robert Miles Jennings, Aanya Simran Malhotra