In digital human modeling, optimization-based methods have been developed to simulate postures and motions due to the ability to predict posture and motion according to different criteria. In robotics area, optimization algorithms have also been used for path planning. Current available optimization methods in either human modeling or robotics are deterministic approaches. However, human anthropometry varies from one person to another and all parameters have uncertainties. Robotics has similar characteristics in terms of interaction with the environment. It is important to take account into these uncertainties in the optimization formulation. Stochastic optimization has been developed to consider uncertainties in the optimization problems. Reliability based design optimization (RBDO) and stochastic programming (SP) address the stochastic optimization problems. RBDO methods include most probable point (MPP)-based approaches and sampling-based approaches. Three policies can be found in the MPP-based approaches: nested RBDO, sing-looped RBDO, and sequential RBDO. For the sampling-based methods, factorial sampling, Monte Carlo sampling, importance sampling, and constraint boundary sampling (CBS) are developed. And some metamodels are often combined in the sampling: response surface models (RSM), kriging model, and artificial neural network (ANN). Three kinds of stochastic programming: stochastic programming with recourse, chance-constrained programming, and stochastic dynamic programming. Stochastic optimization has been applied in many fields in robotics and digital human modeling. This paper attempts to have a literature review on stochastic optimization applications in digital human modeling and robotics.
- Digital human modeling
- Stochastic optimization