Accelerated Gamut Discovery via Massive Parallelization

a) Random (b) NSGA-II (c) LBN-MOBO (ours) Fig. 1.

We propose a highly accelerated approach for discovering the gamut of different design processes. Our method is capable of taking and proposing an extremely large batch of samples at each of its iterations while exploring the design space. In this teaser, we compare the performance of random sampling, NSGA-II (a stochastic multi-objective optimization), and our LBN-MOBO method in identifying the reachable space of a soft robot's tip. All methods operate under the same computational budget of 4 iterations, each with a batch size of 1,000 samples. Our method significantly outperforms existing approaches in both coverage and efficiency.

This paper presents a scalable framework for efficiently discovering the performance gamut of different processes. Gamut boundaries comprise the set of highest-performing solutions within a design space. While sampling methods are often inefficient or prone to premature convergence, Bayesian optimization struggles with taking advantage of existing large-scale parallel computation or experimentation. To address these challenges, we utilize Bayesian neural networks as scalable surrogates for performance prediction and uncertainty estimation. We further introduce a novel acquisition function that combines the diversity-driven exploration of stochastic optimization with the information-efficient exploitation of Bayesian optimization. This enables generating large, high-quality batches of samples. Our approach leverages large batch sizes to reduce the number of iterations needed for optimization. We demonstrate its effectiveness on real-world engineering and robotic problems, achieving faster and more extensive discovery of the performance gamut. Code and data are available at https://gitlab.mpi-klsb.mpg.de/nansari/lbn_mobo.