Inferring Stochastic L-Systems Using a Hybrid Greedy Algorithm

Abstract

Stochastic context-free Lindenmayer systems (S0L-systems) are a formal grammar system that produce sequences of strings based on parallel rewriting rules over a probability distribution. The resulting words can be treated as symbolic instructions to create visual models by simulation software. S0L-system have been used to model different natural and engineered processes. One issue with S0L-systems is the difficulty in determining an S0L-systems to model a process. Current approaches either infer S0L-systems based on aesthetics or rely on a priori expert knowledge. This work introduces PMIT-S0L, a tool for inferring S0L-systems from a sequence of strings generated by a (hidden) L-system, using a greedy algorithm hybridized with search algorithms. PMIT-S0L was evaluated using 3600 procedurally generated S0L-systems and is able to infer the test set with 100% success so long as there are 12 or less rewriting rules in total in the L-system. This makes PMIT-S0L applicable for many practical applications.