Population Dynamics in ARIEL Robotics Systems Featuring Embodied Evolution via Spatial Mating Mechanisms

A team led by Victoria Peterson, Akshat Srivastava, and Raghav Prabhakar introduced a Spatially Embedded Evolutionary Algorithm where robot individuals exist in a physically simulated 2D environment. Using HyperNEAT evolved neural controllers for ARIEL gecko-inspired quadrupeds in the MuJoCo physics engine, they investigated how spatial structure alters evolutionary dynamics. Their experiments compared proximity-based and random mating, finding only a modest 4.9% difference in peak fitness—potentially within stochastic variation. Combining spatial parent selection with stochastic death selection produced unstable population dynamics, while energy-based selection revealed a continuous phase transition between extinction-dominated and explosion-dominated regimes. The density-dependent death selection mechanism achieved 97% completion rates but caused a fitness decline, exposing a fundamental dilemma: decoupled mechanisms create bistable dynamics, positively coupled mechanisms generate counter-selection pressures, and only deterministic fitness-based selection maintains stability.

These findings provide important constraints for future spatial evolutionary algorithm design. The study highlights that spatial structure profoundly impacts evolutionary outcomes, especially when selection pressures are spatially aware. The critical zone count separating extinction and explosion regimes in energy-based selection offers a new lens for understanding population control in embodied evolution. The work underscores the trade-offs between completion rates and fitness optimization, suggesting that deterministic selection may be necessary for stable evolutionary progress. This research has implications for robotics, artificial life, and evolutionary computation, particularly for designing systems where physical robots must evolve in real-world environments.