Scalable Co-Optimization of Morphology and Control in Embodied Machines

Evolution sculpts both the body plans and nervous systems of agents together over time. In contrast, in AI and robotics, a robot’s body plan is usually designed by hand, and control policies are then optimized for that fixed design. The task of simultaneously co-optimizing the morphology and controller of an embodied robot has remained a challenge — as evidenced by the little improvement upon early techniques over the decades since their introduction. Embodied cognition posits that behavior arises from a close coupling between body plan and sensorimotor control, which suggests why co-optimizing these two subsystems is so difficult: most evolutionary changes to morphology tend to adversely impact sensorimotor control, leading to an overall decrease in behavioral performance. Here, we further examine this hypothesis and demonstrate a technique for “morphological innovation protection”, which reduces selection pressure on recently morphologically-changed individuals, thus enabling evolution some time to “readapt” to the new morphology with subsequent control policy mutations. This treatment tends to yield individuals that are significantly more fit than those that existed before the morphological change and increases evolvability. We also show the potential for this method to avoid local optima and show fitness increases further into optimization, as well as the potential for convergence to similar highly fit morphologies across widely varying initial conditions. While this technique is admittedly only the first of many steps that must be taken to achieve scalable optimization of embodied machines, we hope that theoretical insight into the cause of evolutionary stagnation in current methods will help to enable the automation of robot design and behavioral training.


Scalable Co-Optimization of Morphology and Control in Embodied Machines
Nick Cheney, Josh Bongard, Vytas SunSpiral, Hod Lipson