Tao Zhou

Experience

• 

  DIPLOMA University of Applied Sciences

  Feb 2019 - Sept 2019

  Bachelor Thesis

  Bachelor Thesis : Desgin und Konzeption eines autonomen Roboters für die Obsterntes (score: 1.4/1.0)Advisor: Prof. Dr.-Ing. Michael Namokel (Fachbereichsleitung Technik - Vizedekan)

• 

  Technical University of Munich

  Jun 2022 - Mar 2023

  Semester Thesis

  Semester Thesis: A review on instance-level 6D object pose estimation (1.3/1.0)Advisor: Dr. Yinyu Nie (Visual Computing & Artificial Intelligence Lab at TUM)• A review on instance-level 6D object pose estimation• Explore possible avenues for future research : End-to-end Deep Learning-based Method with Refinement for 6D pose estimation

• 

  Volkswagen Commercial Vehicles

  Sept 2023 - Mar 2024

  AI Software Developer Intern for L4 Autonomous Driving | Autonomous Driving MaaS & TaaS (ADMT)

  Designing and Developing AI-Based Remote Guidance for ID. Buzz AD (Autonomous Driving).• Designed and developed an AI-based Remote Guidance (RG) system for Level 4 autonomous vehicles (AV), aimed at providing driving strategies for remote operators to assist the AV in getting unstuck when the Self-Driving System (SDS) in the vehicle can't handle complex scenarios. This innovative system is capable of navigating complex scenarios, thereby improving safety and efficiency in autonomous vehicle operations.• Engineered an AI framework utilizing state-of-the-art artificial intelligence techniques, including large language models such as ChatGPT-4, and reinforcement learning. This advanced framework excels at predicting intricate scenarios and devising recommended maneuvers and safe trajectories, significantly aiding remote operators in guiding Level 4 autonomous vehicles.• Collaborated closely with AMDT U.S. AI software engineers to refine technical solutions, introducing a groundbreaking approach for Disengagement Prediction (Prediction aspect) and Cloud Narrow AI-based Waypoint Guidance ( Motion planning aspect). These contributions, utilizing MPCI and Cloud platforms, are instrumental in augmenting the safety and navigational prowess of autonomous vehicles, ensuring a higher level of support for remote operators. .• Conducted comprehensive data collection and analysis on edge cases to test system robustness. This process involved creating and testing new scenarios, thereby ensuring the system's robustness in real-world applications Show less

• 

  Technical University of Munich

  Apr 2024 - Nov 2024

  Master Thesis: Motion Planning under Uncertainty using POMDP in Autonomous Multi-Agent Racing

  • Developed a motion planning framework based on Partially Observable Markov Decision Processes (POMDP) with an online solver to handle highly dynamic, uncertain, and interactive multi-agent racing scenarios.•Migrated the EPSILON multi-agent simulator from ROS1 to ROS2, ensuring compatibility with the TUM Autonomous Motorsport Simulation platform and improving real-time performance and scalability.•Integrated the updated EPSILON simulator into the TUM Autonomous Motorsport Planning Module, adapting it to generate real-time trajectories for autonomous racing.•Designed and tested various racing scenarios on tracks such as LVMS, optimizing for both safety and efficiency, while minimizing lap times.•Conducted extensive simulations to evaluate the framework's performance, using metrics such as lap time, collision rate, and computational latency.•Validated the motion planning framework against baseline approaches, demonstrating significant improvements in decision-making under uncertainty, particularly in overtaking and obstacle avoidance maneuvers. Show less

• 

  FTM Institute of Automotive Technology TUM

  May 2024 - now

  ADAS Software Developer for the Advanced driver-assistance system (ADAS)

  • Developed key components of the Autonomous Emergency Braking (AEB) system, including object-based collision detection and real-time response integration using ROS2.• Designed and tested Euro NCAP scenarios in CARLA, simulating various dynamic and static obstacles to validate AEB system performance.• Recorded and utilized ROSBAG data from the CARLA Scenario Runner and real autonomous vehicle "EDGAR" to ensure accurate detection and braking responses in both simulated and real-world environments.• Created ROS nodes for calculating and visualizing performance metrics such as Time-to-Collision (TTC) and collision rates, enabling comprehensive evaluation using RViz and ROS2 tools. Show less