Towards self-organized control: Using neural cellular automata to robustly control a cart-pole agent
This work addresses control tasks in robotics or AI by introducing a novel method, but it is incremental as it applies Neural CA to a known benchmark.
The authors tackled the problem of controlling a cart-pole agent by using neural cellular automata (Neural CA) as a differentiable framework, achieving emergent stable behavior for thousands of steps and demonstrating robustness to disruptions like input deletion.
Neural cellular automata (Neural CA) are a recent framework used to model biological phenomena emerging from multicellular organisms. In these systems, artificial neural networks are used as update rules for cellular automata. Neural CA are end-to-end differentiable systems where the parameters of the neural network can be learned to achieve a particular task. In this work, we used neural CA to control a cart-pole agent. The observations of the environment are transmitted in input cells, while the values of output cells are used as a readout of the system. We trained the model using deep-Q learning, where the states of the output cells were used as the Q-value estimates to be optimized. We found that the computing abilities of the cellular automata were maintained over several hundreds of thousands of iterations, producing an emergent stable behavior in the environment it controls for thousands of steps. Moreover, the system demonstrated life-like phenomena such as a developmental phase, regeneration after damage, stability despite a noisy environment, and robustness to unseen disruption such as input deletion.