Junjie ZHANG

Participated in several autonomous driving projects: 1. Deep-Learning based Planning for internal road and garages Developed and implemented deep learning planning model structure for navigation-free scenarios. Designed and deployed data collection triggers to gather user driving data in internal roads for various scenarios. 2. XNGP availability expansion in internal roads and garages Applied a spatio-temporal decision making architecture so as to cooperate with the learning based planning model, enabling rejection of invalid model outputs and selection of high-performing path and trajectory. Developed rule based path and trajectory sampling algorithm, providing diverse and feasible choices for decision making. Designed and implemented a path generation algorithm for indoor garages, ensured high reliability in static and dynamic scenarios, improved overall system robustness. 3. XNGP HD-mapless autonomous driving system Contributed to the real-time mapping developement, focused on extracting and processing key road-level semantic elements under scenarios as road splitting and merging. Developed and implemented lane center lines generation algorithms, enhanced lane topology understanding by specific processing techniques.

Participated in the company’s first 0-to-1 mass-production project for self-developed for Automated Parking Assistance (APA) project: Developed and implemented path planning algorithms, including vertical and parallel parking scenarios, using Hybrid A* algorithm. Optimized computation time by introducing heuristic pruning strategies aligned with human driving habits. Designed and developed a Software-in-Loop (SiL) platform using C++/Qt for offline simulation, validation, and iteration of planning and control algorithms, significantly improving team efficiency. Created an on-board validation platform for real-vehicle testing, enabling visualization, UDP communication with ECUs, and data recording/playback, resolving over 100 real-vehicle issues. Authored functional and performance requirement documents for APA and ultrasonic radars, leading multiple rounds of reviews and establishing a software release gating process to ensure quality. Key Achievements: Successfully delivered the company’s first mass-production APA solution, setting a benchmark for future projects. Improved algorithm iteration efficiency by developing a custom SiL platform, reducing testing and validation time. Enhanced real-vehicle issue resolution by creating an on-board validation platform, enabling faster troubleshooting and analysis.

Keywords: Computer Vision for industrial abrasion analysis, Decision tree of RGB clustering analysis, Python & pyQt for user interface development.

Keywords: Quality analysis and management for universal joints in steering system, VBA tools development.