As a core developer of the Cloud-based Scheduling & Planning System (SPS), was responsible for the end-to-end operations of real-time route planning modules from research, development, validation to integration, deployment and release for the Automated Unmanned Traffic Management (AUTM) System under Beyond Visual Line of Sight (BVLOS) operations.
Delivered critical functionalities including emergency landing and route re-planning, supporting the company to secure China's first BVLOS trial operation license from the Civil Aviation Administration of China (CAAC). And received an exceeded-expectations performance evaluation award.
Dynamic airspace occupation generation based on a spatio-temporal capsule model, include visualization and system-wide conflict computation consistency mechanisms design.
Redesigned a high-performance ST graph framework for long-horizon multi-uav global planning with more precision and more flexibility. Achieved 100x faster conflict-construction (1s ->10ms) and 80% reduction in collision-check (1ms -> 0.2ms) via spatio-temporal decoupling, parallelization, and computational reuse and proposed conflict identification method. Increased system throughput for anomaly handling by 10x (1 drones/s → 10 drones/s), supporting 100+ drone fleets in daily operations.
Improved velocity planning module's performance by optimizing solution space based on binary search, reduced average re-planning route time by 56.1 seconds.
Developed a A*-based heuristic trajectory primitives search algorithm, increased collision safe coefficient by 156% (91.92 -> 234.3) and improved flight smoothness by 10% (0.581 -> 0.526)
Prototyped a scalable TaskFlow-powered concurrent planning pipeline, achieving 2x higher real-time planning capacity per node through parallelized trajectory generation and conflict resolution.
Spearheaded the design and implementation of a ROS-based simulation system from scratch to validate system modules process, benchmark system performance, and accelerate iterative optimization. Executed 100W+ simulated drone flights through multi-scenario batch test, adjustable speed, correctness auto-check and docker-based rapid deployment.
Designed and evolved a modular layered architecture for real-time planning systems, include GRPC interface layer, multi data layer, computation logic layer, ensuring sub-second responsiveness and data consistency.
Led the flight platform selection and function payload development, including lasers and safety switches.
Designed a Android-based Multi-UAV voice operation system: Master-slave arch computing terminal Android Apps, Centralized communication arch via Socket protocol, Human-Machine Interface (HMI) design, Voice command recognition module based on iFLYTEK SDK.
Designed a Multi-UAV collaborative target search algorithm based on human-in-the-loop: Mission-driven hierarchical state machine (HFSM), Customized Voice command instruction sets, Collaborative target dynamic allocation mechanisms.
Developed a Gazebo-based 3D simulations environment for strategy validation and scenario modeling.
Team recruitment and training (8 members), Procurement management, Collaboration management and task coordination, Development sprints and field testing
| Search Algorithms | Collision Avoidance | Reinforcement Learning | Heuristic Optimization | Control Algorithms |
| A* | RRT | ORCA | APF | DQN | MADDPG | GA | PSO | PID |
| Trajectory Generation | Operations Research Optimization | Computer Vision |
| Optimization-based | Analytical-based | ALNS | MIP | MobileNet | YOLO |