A method for determining trajectories of movable physical objects in a space on the basis of sensor data from a plurality of sensors, comprising the following steps:
- a) generating first data objects based on the sensor data, wherein each first data object comprises at least one timestamp and one location specification of an object detected by at least one of the sensors, and a validation function;
- b) providing at least one model object, which represents a parameterization of a trajectory of one of the movable objects with at least one variable parameter;
- for the first data objects:
- c) calculating a correspondence value for an assignment of the data object to at least one of the model objects provided by way of evaluating the validation function of the data object for the model objects;
- d) providing an additional model object and assigning the data object to the additional model object if none of the correspondence values lie in a predetermined range;
- e) assigning the data object to the model object provided that corresponds to the best correspondence value if at least one of the correspondence values lies in the predetermined range;
- and
- f) performing a numerical optimization of a target function for the model object to which the data object was assigned, by varying the at least one variable parameter of the model object, wherein the target function is a function of parameter adaptation functions of at least some of the data objects assigned to the model object, and updating the at least one variable parameter of this model object.