Automated Driving Maneuvers - Trajectory Planning via Convex Optimization in the Model Predictive Control Framework
Doctoral thesis, 2016
Highly automated vehicles have the potential to provide a variety of benefits e.g., decreasing traffic injuries and fatalities while offering people the freedom to choose how to spend their time in their vehicle without jeopardizing the safety of themselves or other traffic participants. For automated vehicles to be successfully commercialized, the safety and reliability of the technology must be guaranteed. A safe and robust trajectory planning algorithm is therefore a key enabling technology to realize an intelligent vehicle system for automated driving that can cope with both normal and high risk driving situations.
This thesis addresses the problem of real-time trajectory planning for smooth and safe automated driving maneuvers in traffic situations where the ego vehicle does not have right-of-way i.e., yielding maneuvers e.g., lane change, roundabout entry, and intersection crossing. The considered problem of generating an appropriate, safe, and smooth trajectory consisting of a sequence of longitudinal and lateral control signals is formulated as convex optimal control problems in the form of Quadratic Programs~(QP) within the Model Predictive Control~(MPC) framework in a manner that allows for reliable, predictable, and robust, real-time implementation on a standard passenger vehicle platform.
The ability of the proposed trajectory planning algorithms to generate appropriate, safe, and smooth trajectories is validated by simulation studies and experiments in a Volvo V60 performing automated lane change maneuvers on a test track. The contribution of this thesis is thereby considered to be a building block for Advanced Driver Assistance Systems (ADAS) regarding yielding maneuvers e.g., lane change, and eventually highly automated vehicles.
Model Predictive Control
Advanced Driver Assistance Systems