Evolutionary Developmental Robotics: Improving Morphology and Control of Physical Robots

Abstract

Evolutionary algorithms have previously been applied to the design of morphology and control of robots. The design space for such tasks can be very complex, which can prevent evolution from efficiently discovering fit solutions. In this article we introduce an evolutionary-developmental (evo-devo) experiment with real-world robots. It allows robots to grow their leg size to simulate ontogenetic morphological changes, and this is the first time that such an experiment has been performed in the physical world. To test diverse robot morphologies, robot legs of variable shapes were generated during the evolutionary process and autonomously built using additive fabrication. We present two cases with evo-devo experiments and one with evolution, and we hypothesize that the addition of a developmental stage can be used within robotics to improve performance. Moreover, our results show that a nonlinear system-environment interaction exists, which explains the nontrivial locomotion patterns observed. In the future, robots will be present in our daily lives, and this work introduces for the first time physical robots that evolve and grow while interacting with the environment.

The trade-off between morphology and control in the co-optimized design of robots

“…Our experiments adopt the robot described in [21], which is a robotic manipulator with a gripper and a glue gun at the end effector. This robot, known as Mother Robot, autonomously assembles children robots by exploring two morphological and three control parameters, which results in a search space of 101,871 combinations.…”

“…Optimization methods, on the other hand, usually do not have biases and solely rely on assumptions concerning the observed data. As a direct example, [21] demonstrates the construction of robots with EA in a model-free approach. Our work contrasts with the EA approach by introducing BO into the MC process.…”

“…This prompts us to rethink how engineering designs are conceived, as morphological traits are usually locked in the very first stages of development. Design methods, such as CAD and FEM, are crucial for morphological predictions, but they are usually disconnected from physical simulations, and thus C happens after M. While an opposite approach would be to optimize an M parameter for a specific C parameter [12], in our previous work [21] we introduced a nested development stage to an EA, where some M parameters evolved while C parameters were frozen, and the improvement of the system depended heavily on which M parameter was chosen to evolve. In here, we show for the first time an analysis of the trade-off between morphology and control of robots in real-life.…”

“…It is undeniable that real-world experiments are needed to ensure the best morphological choice, but robotic morphologies are not easily changeable and design processes with a high number of iterations can slow the process drastically [21]. In a real-world experiment with evolutionary robots [22] a significant improvement in the final MC design could be seen, but the process required the construction of 500 robots to reach the best design.…”

Evolving robot software and hardware

“…With the development of 3D-printing, rapid prototyping, and automated assembly the evolution of robots is becoming feasible, at least in an academic setting [3,13,15,24]. Additionally, researchers have demonstrated how an organism evolved in simulation can be incarnated by using 'wetware' instead of hardware [16].…”