Research on Efficient Heuristic Algorithms for Blocking Flow-shop Scheduling Problem

• Developed a high-performance C++ solver for the Blocking Flow-shop Scheduling Problem (BFSP) utilizing an improved Discrete Invasive Weed Optimization (DIWO) algorithm.
• Engineered a hybrid algorithmic framework by integrating PF-NEH initialization and SRLS local search to enhance solution quality and convergence speed.
• Optimized the neighborhood evaluation function, successfully reducing computational complexity from O(mn²) to O(mn), which significantly accelerated the solving process.
• Validated the algorithm on Taillard’s benchmark instances, achieving superior precision with an Average Relative Percentage Deviation (ARPD) as low as 1.78%.
• Contributed to the design and implementation of core software modules and conducted extensive performance tuning.

[Code]

RISC-V Proxy Operating System Kernel

• Developed a multi-core, concurrency-safe proxy OS kernel based on the RISC-V architecture using the Spike simulator.
• Implemented Sv39 virtual memory with Copy-on-Write (CoW) mechanisms, optimizing memory overhead through page table permission control and page reference counting.
• Engineered process management and scheduling modules (fork, yield, wait) utilizing Round-Robin scheduling and synchronization primitives like semaphores and spinlocks.
• Designed a Virtual File System (VFS) abstraction layer supporting comprehensive file operations, directories, links, and relative path resolution.
• Created an interactive Shell featuring command history and environment variables, while implementing kernel-supported pipes for IPC and dynamic multi-core background task scheduling.

[Code]

RMDB Relational Database Management System

• Implemented SQL lexical and syntax parsing based on Flex/Bison, constructed the AST, and completed semantic analysis including column binding, type checking, and conditional expression validation.
• Built a rule-based query optimizer and a Volcano-model execution engine, supporting operators such as predicate pushdown, projection pushdown, index selection, and join plan generation.
• Designed a page-based storage architecture, buffer pool, and B+ tree composite indexes, supporting equality queries, range scans, and leftmost-prefix matching.
• Implemented transaction rollback and MVCC-based snapshot visibility control, and built a logging and recovery framework to support database crash recovery.

[Code]

Real-time Monitoring Robot System based on YOLOv7

• Managed the collection, cleaning, and preprocessing of multi-class datasets (smoking, fighting, falling, littering) and fine-tuned the YOLOv7 framework to optimize detection accuracy and inference speed for specialized surveillance scenarios.
• Implemented a multi-model object detection system integrated with the SORT tracking algorithm to ensure stable target persistence in complex environments; developed a sliding window algorithm for video stream processing to enable effective recognition of continuous temporal behaviors.
• Designed and deployed an “End-Edge-Cloud” system architecture utilizing the MQTT protocol for efficient data transmission between robots and backend servers; built a comprehensive monitoring client using Dear PyGui to facilitate a seamless workflow from video input and model inference to real-time result visualization.

[Code]