Cyber-Physical System Optimization Techniques in Mechatronic Redundant Accelerometer Array Devices (RAAD)

Abstract

Listen

Translate article

A modeler's approach to optimization greatly improved the calibration of redundant accelerometer array devices (RAADs) for batch estimates of stochastic parameters for zero bias ( $V$ ) and sensitivity ( $V/g$ ). Each device under test (DUT) has an orientation relative to a shared rigid-body frame. The strategy offers cyber-physical systems a reference-free approach for geometry. A function minimizes estimation errors of intrinsic parameters across a gamut of DUTs in array. The algorithmic goal reduces the population average standard deviation $(\overline{\sigma _g})$ estimate of static gravity $(g)$ measurements across more than or equal to three static world poses. RAAD achieves 72 sensor placements in three poses. Comparisons are made to root-mean-square-error ( $\overline{\text{RMSE}}$ ) approaches in previous studies. The optimization model solves a nonlinear system with logical constraints using matlab 's $ {\rm fmincon}()$ function. Precision is achieved for 48 design variables. The RAAD empirical value for $\overline{\sigma _g}$ , initially $0.1042 g$ , improves to $0.0027 g$ and $\overline{\text{RMSE}}$ improves from $\text{0.0282}_{\text{RMSE}}$ to $\text{0.0002}_{\text{RMSE}}$ . The algorithm is tested in simulation with results yielding a high accuracy in estimates for zero bias, $\pm \text{0.35}\%,$ and sensitivity in the range $-0.68\%$ to $ \text{2.46}\%$ . Simulated orientation results are accurate in the range $-\text{0.17}\%$ to $\text{0.61}\%$ across a deterministic grid. Method comparisons indicate that the $\overline{\sigma _g}$ algorithm is superior to $\overline{\text{RMSE}}$ for the calibration of intrinsic parameters.