A fast and reliable hybrid approach for inferring L-systems

Abstract

Lindenmayer systems (L-systems) are a formal grammar system that iteratively create new strings from previous strings by rewriting each of its symbols in parallel according to a set of rewriting rules. The symbols in the string sequence produced can be taken as instructions to produce a visualization of a process over time. They have been especially useful for creating accurate simulations of plants. The L-system inductive inference problem is the problem of inferring an L-system that initially produces a given sequence of strings. Here, a new tool to solve this problem, PMIT-D0L is introduced, that combines projected solutions with linear diophantine equations, heuristics, and genetic algorithm. PMITD0L was validated using 28 previously developed deterministic context-free L-systems of different complexity, and it can infer every L-system in the testbed with 100% success rate in less than 4 seconds, a significant improvement over existing implemented tools.