VCTP: Vehicle-Conditioned Terrain Planning for Off-Road Navigation
研究解决了越野导航中车辆姿态对行驶表面及车身倾斜的影响问题，通过VCTP方法评估不同方向下的地形，并结合D* Lite算法规划最优路径。
0 citationsRead paper
Selected work
Representative Papers
Large-Scale Data Parallelization of Product Quantization and Inverted Indexing Using Dask
This work addresses the high memory and computational costs associated with approximate nearest neighbor (ANN) search in large-scale, high-dimensional datasets. To overcome the limitations of single-machine resources, the authors propose a divide-and-conquer parallelization framework built on Dask that efficiently integrates product quantization (PQ) with inverted indexing. By distributing computation and storage across multiple nodes, the method significantly reduces resource demands while preserving search accuracy. As a result, the computational overhead of large-scale high-dimensional ANN search is brought down to levels comparable to those of medium-scale datasets, enabling scalable and efficient approximate retrieval.
0 citationsRead paper
Characteristic Energy Behavior Profiling of Non-Residential Buildings
Climate change and extreme weather events increasingly threaten the energy infrastructure of U.S. military installations, necessitating robust modeling of non-residential building energy consumption behavior to support resilience planning. Method: This study proposes a machine learning framework integrating multimodal time-series data, combining clustering and predictive algorithms to accurately identify and classify distinct energy consumption patterns; model generalizability is validated using publicly available structured datasets. Contribution/Results: The work establishes the first scalable, baseline energy behavior model specifically designed for military facilities. It enables quantitative assessment of disruption impacts—such as those caused by climate-induced outages—and provides a comparable, reusable benchmark framework for evaluating climate-adaptive resilience measures. By bridging a critical gap in data-driven energy resilience modeling for defense infrastructure, this research advances evidence-based decision-making for mission-critical facility operations under evolving climatic conditions.
0 citationsRead paper
Product Quantization for Surface Soil Similarity
Soil classification traditionally relies on subjective expert judgment and lacks data-driven similarity metrics. To address this, we propose the first machine learning framework integrating Product Quantization (PQ) for soil similarity modeling. Our method jointly embeds high-dimensional spectral and physicochemical features, employs PQ-accelerated approximate nearest neighbor search, conducts systematic parameter-space scanning, and performs sensitivity analysis—thereby overcoming heuristic hyperparameter tuning limitations and enabling interpretable, task-specific soil taxonomy construction. Experiments demonstrate that our approach significantly outperforms both expert-derived classifications and mainstream ML baselines in similarity measurement accuracy, intra-class consistency, and cross-regional generalizability. It establishes a new paradigm for classifying high-dimensional soil data that balances precision with interpretability.
0 citationsRead paper
Recent publications
Latest Papers
VCTP: Vehicle-Conditioned Terrain Planning for Off-Road Navigation
研究解决了越野导航中车辆姿态对行驶表面及车身倾斜的影响问题，通过VCTP方法评估不同方向下的地形，并结合D* Lite算法规划最优路径。
0 citationsRead paper
Large-Scale Data Parallelization of Product Quantization and Inverted Indexing Using Dask
This work addresses the high memory and computational costs associated with approximate nearest neighbor (ANN) search in large-scale, high-dimensional datasets. To overcome the limitations of single-machine resources, the authors propose a divide-and-conquer parallelization framework built on Dask that efficiently integrates product quantization (PQ) with inverted indexing. By distributing computation and storage across multiple nodes, the method significantly reduces resource demands while preserving search accuracy. As a result, the computational overhead of large-scale high-dimensional ANN search is brought down to levels comparable to those of medium-scale datasets, enabling scalable and efficient approximate retrieval.
0 citationsRead paper
Characteristic Energy Behavior Profiling of Non-Residential Buildings
Climate change and extreme weather events increasingly threaten the energy infrastructure of U.S. military installations, necessitating robust modeling of non-residential building energy consumption behavior to support resilience planning. Method: This study proposes a machine learning framework integrating multimodal time-series data, combining clustering and predictive algorithms to accurately identify and classify distinct energy consumption patterns; model generalizability is validated using publicly available structured datasets. Contribution/Results: The work establishes the first scalable, baseline energy behavior model specifically designed for military facilities. It enables quantitative assessment of disruption impacts—such as those caused by climate-induced outages—and provides a comparable, reusable benchmark framework for evaluating climate-adaptive resilience measures. By bridging a critical gap in data-driven energy resilience modeling for defense infrastructure, this research advances evidence-based decision-making for mission-critical facility operations under evolving climatic conditions.
0 citationsRead paper
Product Quantization for Surface Soil Similarity
Soil classification traditionally relies on subjective expert judgment and lacks data-driven similarity metrics. To address this, we propose the first machine learning framework integrating Product Quantization (PQ) for soil similarity modeling. Our method jointly embeds high-dimensional spectral and physicochemical features, employs PQ-accelerated approximate nearest neighbor search, conducts systematic parameter-space scanning, and performs sensitivity analysis—thereby overcoming heuristic hyperparameter tuning limitations and enabling interpretable, task-specific soil taxonomy construction. Experiments demonstrate that our approach significantly outperforms both expert-derived classifications and mainstream ML baselines in similarity measurement accuracy, intra-class consistency, and cross-regional generalizability. It establishes a new paradigm for classifying high-dimensional soil data that balances precision with interpretability.
0 citationsRead paper