Module morphology
This project investigates how the number and position of available connection faces in a module influence the evolvability of the modular robot. Based on EMERGE, we defined the number of connection faces and their relative positions as morphological parameters. Afterwards, we evolved the morphology and control of robots composed of different morphologies of EMERGE modules in a robotic simulation platform, and then transfered the simulated robots to the real robots.
We classify the morphology of connection faces into five types. In addition, each module type has four possible orientations as the connector allows us to connect two faces after a 90 degrees rotation.
We evolved the morphology and control of the modular robot assembled by the different morphologies of EMERGE module using the Edhmor system [1] and V-REP simulator. The best morphologies of the modular robot assembled respectively by these five module types were obtained in the simulator, and then transferred them into the real robots.
Because the evolutionary data is not normally distributed, we should use Mann-Whitney U tests to check if there is the statistically significance difference between these five module types. If it is, then refer to the statisticall indicators such as mean and midean fitness to show which module type is the best.
In the boxplot figure, there is the statistically significant difference beween type 4 and other module types. In addtion, the mean and median fitness of type 4 is both better than other types. Thus, type 4 module type can lead to the best robot morphologies for the locomotive task.
[1] The source code of Edhmor system is available at: https://bitbucket.org/afaina/edhmor