In short, it's essentially a topology-based approach to solving a classical optimization problem. Quite novel, really, though it's not clear if the solution offers a whole lot of benefit over brute force calculation in terms of saved computation time. The paper is here.
What struck me as particularly interesting was the authors' proposal to physically implement their "blob" model:
...it would be satisfying if this virtual material could be implemented and embodied in a real physical substrate with the desired physical (for example visco-elastic, free energy minimisation) properties.