generate_continuous_values(executor, state, nr_iterations=1, nr_samples=5)[source]#

The function generate_continuous_values and expand_continuous_values jointly implements the incremental search algorithm for Task and Motion Planning.

Basically, the algorithm starts from generating a large collection of possible continuous parameters by “expanding” from the continuous parameters in the input state. Next, it reduces the TAMP problem into a basic discrete search problem. The downside of this approach is that it requires grounding a large collection of possible values, but it is in theory probabilistically complete.

  • executor (PDSketchExecutor) – a pdsketch expression executor.

  • state (TensorState) – the input state.

  • nr_iterations (int) – the number of iterations.

  • nr_samples (int) – the number of samples for each generator.


a dictionary of continuous values.

Return type:

Dict[str, List[TensorValue]]