Publications & Patents
See Research Experience for details.
Publications
Vision-And-Language Navigation for Unmanned Systems: Progress and Perspectives
Dian Shao†, Like Liu, Zhengzheng Xu, Junqiang Bai
Proceedings of the 5th International Conference on Autonomous Unmanned Systems (ICAUS 2025)
TL;DR: This survey traces the evolution of Vision-and-Language Navigation (VLN) from indoor to outdoor/aerial settings, emphasizing how foundation models (VLMs/LLMs) enable zero-shot, language-guided autonomy through open-vocabulary perception and reasoning, while reviewing benchmarks, methods, challenges, and future directions for intelligent embodied agents.
Knowing the Self, Understanding the World: A Dual-Cognition Benchmark for UAV Spatio-temporal Reasoning with MLLMs
Like Liu, Haitao He, Zhengzheng Xu, Hongzhe Li, Shuchang Zhang, Dian Shao†
Submitted to ACM MM 2026 Dataset Track
TL;DR: We introduce UAV-DualCog, a new benchmark for evaluating multimodal LLMs on aerial dual-cognition tasks (reasoning about both UAV self-state and environment state in multiview spatio-temporal contexts) and show that current lightweight MLLMs struggle significantly, especially with self-state understanding.
Contributions: 1. Led the project and designed the tasks. 2. Developed the benchmark construction pipeline and implemented experiments. 3. Visualization and illustration during paper writing.
FineCog-Nav: Integrating Fine-grained Cognitive Modules for Zero-shot Multimodal UAV Navigation
Dian Shao†, Zhengzheng Xu, Peiyang Wang, Like Liu, Yule Wang, Jieqi Shi, Jing Huo
Findings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2026 (CVPR 2026 Findings)
TL;DR: We propose FineCog-Nav, a cognitive-inspired, modular zero-shot framework for UAV vision-language navigation, where specialized moderate-sized models handle distinct reasoning functions; evaluated on our new fine-grained benchmark AerialVLN-Fine, it outperforms existing baselines in instruction following, long-horizon planning, and generalization.
Contributions: 1. Performed quality filtering on AerialVLN and refined referring expressions, producing 300 annotated instruction-trajectory pairs, forming fine-grained dataset AerialVLN-Fine. 2. Designed and developed the human study. 3. Visualization and illustration during paper writing.
FineTec: Fine-Grained Action Recognition under Temporal Corruption via Skeleton Decomposition and Sequence Completion
Dian Shao†, Mingfei Shi, Like Liu
Proceedings of the 40th AAAI Conference on Artificial Intelligence (AAAI 2026)
TL;DR: We propose FineTec, a robust framework for fine-grained action recognition from temporally corrupted skeleton sequences, which combines context-aware skeleton completion, semantic spatial decomposition, physics-driven acceleration estimation, and GCN-based fusion—achieving state-of-the-art performance on both coarse- and fine-grained benchmarks under severe data corruption.
Contributions: 1. Led the annotations for 12K segments of competitive sports videos based on Gym99, forming a extended fine-grained skeleton posture dataset Gym288-skeleton. 2. Implemented the skeleton sequence completion module using in-context learning and multi-mask strategies. 3. Visualization and illustration during paper writing.
Intelligent UAVs Empowered by Large Models: Progress, Applications and Perspectives
Dian Shao†, Chu Tang, Min Chang, Like Liu, Yule Wang, Hao Li, Junqiang Bai
Acta Aeronautica et Astronautica Sinica
TL;DR: This survey explores how Large Foundation Models (LFMs) are transforming UAV intelligence—covering architectures, core function enhancements (perception to interaction), high-level cognitive capabilities (reasoning, memory, etc.), applications in key decision-making tasks, and challenges in safety, deployment, and future directions toward efficient, cognition-aware, and collaborative aerial systems.
Whole-Body Control for Multi-Contact Balancing of Humanoid Robots: Design and Experiments
Bernd Henze (Author), Liang He (Translator), Yan Ma (Translator), Like Liu (Translator), Yuting Shi (Translator)
National Defense Industry Press, Beijing
TL;DR: Translated monograph on whole-body control for humanoid robots.
Patents
Fine-grained Action Recognition under Temporally Corrupted Sequences via Skeleton Decomposition and Sequence Completion
Dian Shao, Mingfei Shi, Like Liu
Granted, No. ZL202511913688.7
TL;DR: Invention patent for fine-grained action recognition under temporal corruption via skeleton decomposition and sequence completion.
Zero-shot Vision-and-Language Navigation Method for UAVs via Integration of Fine-grained Cognitive Modules
Dian Shao, Zhengzheng Xu, Peiyang Wang, Like Liu, Jieqi Shi, Yule Wang, Jing Huo
Published, No. CN121230728A
TL;DR: Patent for zero-finetuning UAV vision-language navigation using cognitive modules.
Humanoid Robot Navigation Method Using a Hybrid Fuzzy Neural Network and Adaptive PID Control Algorithm
Like Liu, Boyang Sun, Liang He
Published, No. CN120385341A
TL;DR: Humanoid robot navigation method combining fuzzy neural network and adaptive PID control.
Fine-grained Human Motion Generation Method Based on Physical Laws and Skeleton Guidance
Dian Shao, Mingfei Shi, Junqiang Bai, Shengda Xu, Haodong Chen, Yongle Huang, Like Liu
Granted, No. ZL202510575275.6
TL;DR: Invention patent for fine-grained human motion generation using physical laws and skeleton guidance.
GTD-based Task Assignment and Management System
Like Liu, Shuchang Zhang, Zhenguo Li, et al.
No. 15861786
TL;DR: Software copyright for GTD-based task assignment and management system.
UAV-enabled Low-altitude Logistics Planning System
Yutong Li, Like Liu, Haoxuan Li
No. 13950938
TL;DR: Software copyright for UAV-enabled low-altitude logistics planning system.
Time-window Constrained Multi-node Path Planning System
Like Liu, Haoxuan Li
No. 12125321
TL;DR: Software copyright for time-window constrained multi-node path planning system.