Research On Merging Behavior Of Autonomous Vehicles Considering Collision Avoidance Intention

Ensuring the safety of mixed traffic conditions between autonomous and conventional vehicles is a crucial area of research in transportation.As autonomous driving technology becomes increasingly prevalent on highways,ensuring the safety of mixed traffic situations will become even more important.Previous studies have shown that merging sections on highways experience higher traffic delays and accident rates,with perception and decision errors often being the root cause of these accidents.In the merging scenario of mixed traffic state,there are two kinds of research subjects:autonomous merging vehicles and traditional conflict vehicles in the target lane.When the conflict vehicles observe the existence of merging vehicles in the acceleration lane,they may use the way of lane change to avoid,at this time,if the merging vehicles can recognize the intention of conflict vehicles to change lane in time,they can make the merging decision in advance to improve the efficiency of merging.In the connected car environment,the self-driving vehicle can obtain the location and speed information of the conflicting vehicles in the target lane in real time,and determine the driving intention of the conflicting vehicles through the historical time series trajectory,and the intention recognition result will have a direct impact on the decision-making process of the self-driving vehicle.At the same time,as a small proportion of traffic participants in the current mixed traffic environment,the merging behavior of self-driving vehicles should be in line with the psychological expectations of human drivers.To address the above issues,this thesis proposes a dynamic acceptance gap merging strategy that considers the avoidance intention of conflicting vehicles and conducts a simulation study on self-driving vehicles in the highway merging scenario.Firstly,trajectory data of converging and nearby vehicles on the I-80 section of the NGSIM open source database were extracted.Segmented cubic polynomial interpolation and Savitsky-Goley filtering were applied to reject data outliers and reduce noise,and the processed trajectory data were used as input to a hybrid Gaussian Hidden Markov(GMM-HMM)model for driving intention classification,specifically lane change avoidance and lane keeping.The GMM-HMM model was validated to be more than 90%accurate in recognizing lane change intentions using test data.Next,a braking safety model was constructed and analyzed in the merging scenario,which led to the definition of the braking safety distance of the acceleration lane and the following car after merging.Based on gap acceptance and loss aversion theories,the optimal merging gap to maintain following safety in the no-urgency state and the minimum merging gap to ensure obstacle avoidance safety in the urgency state were defined.A dynamic acceptance gap model was developed and compared with the traditional single reference standard merging gap model,which showed that the dynamic acceptance gap model is more consistent with the psychological process of human driving decision-making while ensuring safety.Using the dynamic acceptance gap model,a benefit function was established to determine the longitudinal trajectory of the merging vehicle in the acceleration lane,which includes merging possibility,merging efficiency,and acceleration comfort.Conflicting vehicle intention recognition results were combined to analyze the process of autonomous vehicle merging strategy execution.Finally,based on the vehicle kinematic model and trajectory constraints,the quintuple polynomial convergence curves are solved for 15m/s,20m/s,25m/s vehicle speed,and 1m/s~2,2m/s~2,3m/s~2acceleration,respectively.The joint simulation of Pre Scan and Matlab/Sinmulink is used to simulate and verify the analysis of different intention scenarios of conflicting vehicles under far and near distance conditions,respectively.The simulation test results prove that the driving intention recognition model proposed in this paper can respond to the conflicting vehicles’intention to change lanes within 0.3s in real time,and the self-driving merging strategy based on the dynamic acceptance gap model can reduce the impact of merging behavior on traditional conflicting vehicles while ensuring driving safety.The final proof is that the merging decision model in this paper has good effectiveness in the state of mixed driving and traditional vehicles.