Data Figure 3Data to reproduce Figure 3, ternary plots showing group performance of handcoded controller in flat and sloped environment. The first field indicates the length of the sloped part of the environment (6 meters for the sloped environment, 0 for the flat), the following three indicate the number of robots engaging in the three strategies, the fifth field indicates the fitness and the last field the number of items dropped in the cache.dataFigure3.zipData Figure 4Data to reproduce Figure 4, showing group performance and amount of task partitioning over subsequent generations for each of the 22 evolutionary runs. The first three fields indicate the ID of the evolutionary run, generation and repetition (respectively), the fourth and fifth fields indicate the number items retrieved in a partitioned and non-partitioned way (respectively, the sum of which row-wise corresponds to fitness), and the sixth indicates the average absolute robot speed projected along the main axis of the environment.dataFigure4.zipData Figure 5Data to reproduce Figure 5, robot densities and trajectories for a typical run. The first fields indicates the timestep, the second, third, fourth and fifth fields indicate the x-axis (main environment axis) coordinate of the four robots, and the last field contains the number of items present in the cache.dataFigure5.zipData Figure 6Data to reproduce Figure 6, showing the effect of the degree of task specialization and average linear speed on the fitness performance of the 22 controllers evolved from first principles. The first and second fields indicate the ID of the evolutionary run and of the repetition, the third and fourth fields indicate the number of items retrieved in a partitioned and non-partitioned way (respectively), the fifth field indicates the proportion of items retrieved in a task partitioned way, the sixth and seventh fields indicate the average linear speed and the same quantity as a proportion of the maximum theoretical speed, and the last field indicates the fitness.dataFigure6.zip

Division of labor is ubiquitous in biological systems, as evidenced by various forms of complex task specialization observed in both animal societies and multicellular organisms. Although clearly adaptive, the way in which division of labor first evolved remains enigmatic, as it requires the simultaneous co-occurrence of several complex traits to achieve the required degree of coordination. Recently, evolutionary swarm robotics has emerged as an excellent test bed to study the evolution of coordinated group-level behavior. Here we use this framework for the first time to study the evolutionary origin of behavioral task specialization among groups of identical robots. The scenario we study involves an advanced form of division of labor, common in insect societies and known as “task partitioning”, whereby two sets of tasks have to be carried out in sequence by different individuals. Our results show that task partitioning is favored whenever the environment has features that, when exploited, reduce switching costs and increase the net efficiency of the group, and that an optimal mix of task specialists is achieved most readily when the behavioral repertoires aimed at carrying out the different subtasks are available as pre-adapted building blocks. Nevertheless, we also show for the first time that self-organized task specialization could be evolved entirely from scratch, starting only from basic, low-level behavioral primitives, using a nature-inspired evolutionary method known as Grammatical Evolution. Remarkably, division of labor was achieved merely by selecting on overall group performance, and without providing any prior information on how the global object retrieval task was best divided into smaller subtasks. We discuss the potential of our method for engineering adaptively behaving robot swarms and interpret our results in relation to the likely path that nature took to evolve complex sociality and task specialization.