Start with lots of real stories of motion in a horizontal plane, mostly qualitative and semi-quantitative. These could be a 'drone's-eye' (perhaps even live) or 'bird's eye' view: a two-dimensional view of a tracked motion. These have exploitable resonances in lived-in world representations, such as GPS-enabled mappings of personal journeys (Strava, MapmyRide), maybe aerial traffic cams or city mapping of traffic flows. Start with stories of journeys told in words and diagrams with very young children, so rehearsing the need for taking a point of view. Try switching between first-person and third-person perspectives, and maybe exploiting links to local approaches to geometry, probably most often met in varieties of turtle geometry. Encourage framing of first and third-person descriptions using 'how fast, for how long and in what direction' or 'how far, and in what direction'. This lays in a store of resources to draw on when predicting displacements from velocities later on.