i am using world tracking configuration and .gravityandheading . I locate 3d castle and guards in ar. I do that first detecting image and locate animated object help of location of image. but when I turn camera and move phone or iPad. objects misses their locations. they are moving awkwardly. I find temporary solution by detecting image again(i added objects to -renderer didadd node-'s node ). when I detect image again objects goes their own position but they misses again. maybe not much big but sometimes it goes up to 1 meter long. how can I optimize positions. we want to show users continuos experience when wandering around. but missing positions big problem.