| Promot ing by advanced techno logies such as art ificia l intellige nce and 5G,in order to improve our countr y’s transportation status and enhance our nationa l scie nt ific and techno logica l inno vat ion capabilit ies,many cross-enterprise consort ia exp loring driver less ve hic les have appeared at home and abroad.In view o f t he proble ms o f sma ll parking Spaces and comp lex parking environme nt on urban roads,to relie ve the operating pressure of novice drivers and ensure parking effic ienc y and safety,this paper pursue a further research on the content of automat ic parking and obstacle avoidance in dynamic interactive environment.Aiming at the decis ion-mak ing d iffic ulty o f avo id ing obstacle by adjust ing ve locit y profile or path profile in dynamic interact ive parking scenar ios,a long-term trajectory collaborative decis ion-making met hod for auto mat ic parking was proposed in this paper.By cons idering the obstacle behavior,ve hicle k inemat ics model and d irected space-time map were establis hed to improve t he child node expans ion met hod.Comb ined w it h Reeds-Shepp path pla nning met hod,the effic ienc y he ur ist ic funct io n is proposed by foreward velocity integrat ion Met hod.Based on t he artific ia l potent ia l fie ld met hod,the dyna mic obstacle r isk potent ia l fie ld mode l was built to describe the safety loss of parking trajectory by introducing the speed,size and type factor of the obstacle.Based on the hybr id A-star search framework,a mult i-objective search algorit hm cons idering path length,t ime effic ie ncy,safety a nd c omfort was designed.And the cooperative decis io n was compared with the hybr id A-star path to ana lyze the trajectory efficiency and feasibility.Due to the roughness of t he init ia l so lut ion and easy deviat ion o f long-term trajectory of obstacles,a short-ter m trajectory optimizer for automat ic park ing syste m was designed based on the nonlinear model predict ion met hod(NMPC).At the same time,comb ined w ith the global minimum deviatio n target and the loca l collis io n avoidance target,the veloc ity and the front w heel angle profiles are simultaneous ly solved to ensure t he global a nd loca l optima lit y.Further more,the mult ip le shot d iscret izat io n of t he AC ADO solut ion fra mework,the genera lized Gauss-Newton proble m approximat ion,the real-time iterat ion(RTI)acceleratio n strategy,and the C la ngua ge comp ilat ion struct ure are adopted to accelerate the solut ion effic ien c y of NMPC.C omparing the optimized trajectory wit h the results of sp line opt imizat io n,it is found that t he trajectory pla nned by t he met hod i n this paper is s mooth and safe,which can meet the require ments o f real-time calc ulat ion and provide more accurate reference information for subsequent tracking.Taking the interference caused by the veloc ity change into account,the latera l tracking and lo ngit ud ina l tracking are decoupled.The hierarchica l control met hod is adopted in t he longit udina l track ing strategy,in the upper layer,the doub le PID tracking strategy is adopted.In the lower layer,ve locit y,accelerat ion and thrott le brake calibratio n tables are designed based on table lookup met hod.In the latera l tracking strategy,In the latera l track ing strategy,to avo id the vehic le escapes the preset path in the case of driver interve nt ion,based on the s liding mode met hod,the non-time parameters are used as variables to analyze the tracking deviation.In addit ion,for the jitter proble m caused by the exponent ia l approach law,filtering and hyperbolic tangent funct ion are used to improve it.C ompar ing the method wit h the tradit iona l exponent ial approach slid ing mode track ing control strategy in the Simulink & C arsim e nviro nment,t he accuracy of t he tracking met hod in t his paper is verified.Fina lly,in order to test the robust ness of the trajectory decis ion-mak ing and tracking method proposed in this paper,four common human-vehic le interactive parking scenar ios are built in t he Prescan environme nt,and the vehic le model in C arsim and the pla nning contro l model in Matlab/Simulink are joint ly simulated.The results show that the met hod in this paper can successfully complete the parking task by adjust ing its own prior ity in the four scenar ios,and has better trajectory decision-making ability and tracking effect. |
