MCity Autonomous Vehicle Challenge

Project Overview: Designed and deployed a robust Model Predictive Controller (MPC) for a high-fidelity, city-scale autonomous driving simulation. The challenge involved navigating complex urban scenarios with diverse background agents and real-world crash data constraints.

Key Contributions:

• Robust Control Logic: Engineered an MPC capable of handling aggressive competitive environments, prioritizing safety without sacrificing progress.

• Performance Metrics: Achieved a 99.39% Traffic Rule Compliance rate (Top 3 ranking) and a 93.33% Trajectory Completion rate (2nd highest overall), significantly outperforming baseline algorithms in obstacle avoidance and lane-keeping.

• Simulation & Testing: Validated the stack through continuous testing loops in Mcity’s virtual proving ground, iterating on safety and comfort scores based on dynamic feedback.