Automatic Decision Making and Planning at Crossroad

The project aims at designing an automatic decision making system for intelligent vehicle at a typical crossroad. My work focuses on decision making module design and implementation (by C++), mainly including behavior recognition of surrounding vehicles, decision of local driving target, and trajectory planning. Surrounding vehicles behavior recognition is achieved by a HMM model, which defines vehicle’s behavior as hidden state and infers probability distribution of behaviors every time step. Local driving target is decided by carefully designed rules, and its result is used for trajectory planning, which is realized by an algorithm based on DAG and A*.