Research On Eco-driving Control Strategies For Autonomous Driving Vehicle

The autonomous driving vehicle has great potential in reducing the fuel consumption,guaranteeing driving safety and improving transportation efficiency,simultaneously.However,it has strong physical dynamics,high real-time requirement and complicated constrains,making its control problem.In this thesis,an Eco-driving control strategy is proposed,combining fuel-saving trajectory planning algorithm and driving behavior improving algorithm,aiming at solving the fuel consumption optimizing problem for an autonomous passenger vehicle with a gasoline engine.First,to solve the fuel-saving path planning problem,a trajectory prediction algorithm based on driving intention is proposed,which can obtain the safe driving area for the autonomous vehicle.A prediction algorithm containing a vehicle motion model with lane line data in high-definition map(HD-map)correction is proposed.This algorithm expands the prediction time window and enhances the prediction accuracy.Besides,A dynamic programming method is utilized in path planning algorithm based on non-uniform space sampling strategy,in which a path planning cost function consists of fuel consumption,vehicle safety,vehicle dynamics and driving comfort.The cost function make the fuel consumption importance in path planning.For the fuel-saving speed planning problem,a convex speed panning algorithm is proposed,which containing a fuel consumption cost function based on vehicle fuel consumption model.The multiple target speed optimization is achieved through tuning the cost function weight coefficients,which contains a fuel consumption cost based on a vehicle fuel model.The effectiveness of trajectory planning algorithm is validated in Prescan-Simulink-Car Sim simulation platform.The simulation results demonstrate that a 28% fuel consumption reduction can be achieved by using the proposed speed planning algorithm.On the other hand,for the purpose of economically and accurately tracking the desired path and speed,an active disturbance rejection-based controller is proposed,combining the vehicle model-based feedforward controller and disturbance rejection controller.A longitudinal vehicle model containing the mean value engine model and a lateral vehicle model based on steady-state steering hypothesis are established,with an extended state observer(ESO)compensating the uncertainties from vehicle model inaccuracy and environment disturbances.This thesis constructs a driving behavior space using maximal accelerator pedal depth and accelerator positive rate,and a particle swarm optimization(PSO)algorithm is utilized to search the optimal value in driving behavior space during the process of vehicle driving.The Car Sim simulation results demonstrate that the proposed speed controller has a 1.9% fuel saving compared with Car Sim speed controller,and there will be another 2.37% fuel reduction after the driving behavior optimization.Finally,an autonomous driving prototype vehicle is constructed.The trajectory prediction,trajectory planning and vehicle control software module are developed using Python,and they are running in the XAVIER on-board computer.Then,the Eco-driving control strategy is validated at Tianjin University’s autonomous driving test field.Experiment results show that the runtime of trajectory planning and control algorithm is less than 100 ms,and the fuel consumption results show that trajectory planning algorithm can make a 11.3% reduction of fuel.Otherwise,the optimal fuel consumption-autonomous driving case can achieve a 10.4% fuel reduction compared to an aggressive human driver and a 4.9% fuel reduction compared to a gentle human driver.