Situated Design Optimization of Haptic Devices

Abstract

It is a complex task to develop and optimize a high-performing haptic device. Design optimization scenarios with predefined and fixed sets of performance requirements are presented in literature. However, the early design optimization phases for haptic devices are characterized by requirement conflicting requirements with uncertainties. With a lack of knowledge, and/or an ill-defined design problem, the challenges are not only to find a high quality solution with reasonable computational effort. In this paper, a previously proposed model-based framework and methodology for multi-disciplinary design optimization of haptic devices is further developed to enable situated design scenarios, i.e. design cases that may be characterized by changing requirements, constraints, and/or performance objectives, driven by the knowledge gained in the design and optimization process itself. To provide both precision and computational efficiency, the proposed situated, i.e. flexible and adaptable, framework is based on an approach that combines design-of-experiments (DOE) with meta-modelling methods for multi-objective optimization problems. The proposed methodology is described and verified with a 6 degree-of-freedom (DOF) TAU haptic device optimization scenario, with changing ranges for the design variables and the constraints. Results from the case study strongly indicate that a thoroughly balanced and sequential DOE and metamodelling process is capable of being both effective and efficient in a situated design scenario. It is shown that the knowledge gained in the process, e.g. the number of sampling points and the most appropriate training method, may be used to efficiently balance the required computational effort with the required level of accuracy.