A Position Control-based Approach to Stiff Objects Haptic Rendering
Time: Tue 2021-06-08 10.00
Subject area: Machine Design
Doctoral student: Yang Wang , Maskinelement, Maskinkonstruktion (Avd.)
Opponent: Professor Anders Robertsson, Lund University
Supervisor: Professor Kjell Andersson, Mekatronik, Maskinkonstruktion, Maskinelement, Maskinkonstruktion (Avd.); Lei Feng, Maskinkonstruktion (Avd.), Inbyggda styrsystem, KTH-centrum inom inbyggda system, ICES, Mekatronik, Maskinelement
With electronic components and computational power becoming more portable and available to general consumers, applications like virtual reality (VR) and mixed reality (MR) are on the rise and renovating the industry of training simulators. Though it is not fully accepted in surgical robots, haptic force feedback can be safely introduced into various surgical training simulators to convey the necessary haptic cues needed to develop hand-eye coordination and motor skills. However, with the current generation of haptic device hardware, it is still challenging to render high stiffness virtual objects while stability allows. Passivity-based approaches are commonly used to ensure stability but they may lead to conservative behavior. Therefore, new approaches are needed to explore other ways of rendering high stiffness without losing stability. This thesis proposes a position control-based haptic rendering approach that designs a position controller to do force computation for the rendering of high-stiffness objects. Instead of computing a penalty force signal that mimics the desired interaction and softly constrains the objects to have little penetration, this proposed approach aims to achieve no penetration with the help of a position controller and sensors. The approach can be easily adapted and applied on any haptic device since it relies on the building of an implicit simulation model which can be linearized at one or several operating points to describe the full dynamics of the system. Different user scenarios, disturbance signals, and evaluation metrics are also proposed to assess and compare the performance with other common approaches. The results show that the proposed approach can stably render a virtual wall with stiffness higher than the common way of modeling the wall as a spring-damper system. By using a piece-wise linear model generated based on multiple operating points and a gain scheduling controller with linear interpolation, the performance of the system is further improved.