I have a numpy array of shape n x m x r, where the n axis represents an object, the m axis represents a timestep and the r axis represents a position vector in 3-d space. I have an array containing three (x, y and z position values) of m objects at n points in time. This is the format my data is delivered in (from the python module sgp4, specifically a Satrec_Array if anyone's interested) so I can't move this further up the data processing chain. I want to be able to represent this as an m x n array of position vectors, so essentially "collapsing" the position axis into a single array, so an array containing 20 elements, each of which is a position vector relating to an object at a time. I'm struggling to find a way to do this efficiently - I can do it with bog standard python loops, but as I scale up the number of objects and timesteps this becomes massively inefficient. I'm not very well versed in numpy, but none of the solutions I've searched or attempts at using methods such as [h/v/d]stack etc. have given me the right final shape. I also looked into vectorization but as far as I can see that just implements a loop under the hood? Can anyone help? Continue reading...
