Trajectory planning for antagonistic non-linearly quantized SMA actuator arrays based on evolutionary optimization

This paper presents an evolutionary optimization approach to the control of antagonistic shape memory alloy (SMA) actuator arrays. The actuator and control architectures are inspired by the neuromuscular system, which achieves smooth and accurate movements despite nonlinearities and uncertainties in control and estimation. SMA actuators produce contractile forces and have a high power-density, but exhibit nonlinearity and significant hysteresis in their thermo-mechanical behavior. A floating-point quantization (FPQ) schema is used to determine a non-uniform array distribution, resulting in motor variability similar to that observed in humans. Actuator arrays are used in an antagonistic configuration for joint control of a planar link. An evolutionary optimization approach is used to determine the optimal control inputs for point-to-point trajectories.