Mobile Phone 3D Crowd Animation Autogeneration

Lu Ruqian,who is an academician of Chinese Academy of Sciences,proposed an animation creating technology which is created by the assisting of computer in the full creating life with artificial intelligence technology in nineties of last centry,and an animation autogeneration system Tiane which can generate animations just by inputing story texts without any artificial operatings is designed and realized by him and his teams,the Tiane system can improve animation creating efficiency greatly.Zhang Songmao,who is a researcher of Chinese Academy of Sciences,proposed applying this technology to SMS,and then the full life computer assisted mobile 3D animation autogeneration system(the following abbreviations mobile 3D animation system)which can generate an animation vedio can be played on mobile phone by only inputing a SMS was designed and realized by our laboratory with the lead of her.Crowd animation is usually used to simulate the motion behaviors of a crowd of animals,people or other things in a specific environment.Its application area includes movies,games,robots,safety and so on.This paper proposed to realize a crowd animation autogeneration system in the mobile 3D animation autogeneration system to enrich the expression form of autogenerated animations.The main work of this paper includes:First,design and realize the crowd animation qualitative planning module.A crowd ontology base used to describe the classes of agent(individual),crowd motion space and crowd motion behavior is built.And the semantic relationships with the ontology base which is already designed in the mobile 3D animation system are constructed.A crowd rule base is constructed to build the realtionships among different classes in the ontology base,and is inferenced by a rule engine with the information of the topics and templates of SMS as well as the selected animation scene,then the suitable agent’s name,crowd motion space,crowd motion behavior and some other information are output and used by the crowd quantitative computing module.Second,design and realize the crowd animation quantitative computing module.A markov decision process is used to model the behavior of single agent,then the single agent markov decision process model is generalized to multi-agents,crowd aim position and start positions selecting algorithm,collision avoidance strategy,muti-agents motion strategies and trajectories autogenerating algorithms are designed.Comparing with other crowd animation realizing methods,the aim position and start positions selecting algorithm designed in this paper can be applied in different scene space,and can be correspond to crowd behaviors;the computation efficiency of planning the next action for agent by the motion strategy designed in this paper is higher when the amount of agents is large;the method for estimating the whole timesteps needed to accomplish the crowd behavior can be used to determine appropriate frames for each timestep and make the crowd behavior accomplished near the end of the animation.Open SMS are used to evaluate the system,for the feasibility experiment,the percentage of SMS which successfully generated crowd animation is 94.81%;for the diversity experiment,the qualitative planning results are all equal to or more than 7 when each SMS is tested 10 times repeatedly;for the ornamental value experiments,average score is 6.99(full score is 10)by 51 questionaires.This is the first time to design and realize the crowd animation autogeneration function in the mobile 3D animation system.It enables the system to express the content of SMS by crowd behavior and enriches the expression form of animations to some degree.