Co-Operative Highway Lane Merge Control of Autonomous Self-Driving Ground Vehicles

ADAS (Advanced Driver Assist System), promises to reduce automotive accident deaths, improving road safety as well as enhance the driving experience, where the driver can work or entertain instead of paying attention to the road. It also helps to reduced cost of taxi services (75% of taxi cost is the cost of human driver) enabling elderly and low-income population to commute easily. Autonomous driving technology investment has increased at an accelerated pace in recent years, reaching close to $100B per year as of 2017. Today’s ADAS technology is more advanced and is picking up its pace to replicate many of the existing human driving maneuvers that a human encounter on the road. One such maneuver is the lane change and merge, where a self-driving vehicle in its autonomous mode is required to perform lane change and merge motions safely. The lane change and merge maneuver can arise in different situations like when overtaking a vehicle within the highway lanes, or when a lane ends or when the vehicle is performing a highway merge where the vehicle is moving from an adjacent local lane to the main highway lane. Unlike, simple lane change or overtaking on highway lanes, the highway merge is challenging as it involves different speed limits on each lane especially in the absence of a parallel acceleration lane. Moreover, it requires high coordination between vehicles to safety implement the maneuver respecting traffic laws and conditions. This dissertation proposes a new co-operative nonlinear model predictive optimal controller to solve the highway lane merge problem for connected autonomous vehicles. The nonlinear model prediction formulation uses a kinematic vehicle model for the vehicles involved in the highway merge. The optimal control formulation is solved using simultaneous direct multi-shooting method. The formulation minimizes the cost of a highway merge problem considering the tracking error, driver comfort and safety considerations while subjected to various constraints such as steering angle bounds, vehicle speed limits and acceleration limits. Simulations are first performed to demonstrate the nonlinear model predictive control-based optimization for a single vehicle tracking a local merge path to the highway merge path subjected to state and control constraints. For simulation purpose, the optimization controller is exported to C-language using an open-source software environment and algorithm development ACADO toolkit. The exported code is used in a MATLAB® based simulation and tested for tracking the merge profile. Then, the formulation is extended further, to demonstrate an implementation of a co-operative based nonlinear model predictive optimization scheme for two vehicles. The controller scheme results are presented based on a highway merge profile as seen on Dwight D. Eisenhower Expressway eastbound, from S. Central Avenue, Chicago, Illinois. Simulation results based on two connected vehicles are demonstrated for various scenarios that include regular traffic conditions, heavy traffic conditions and different vehicles leading the highway merge.