Real-Time Point-Cloud-Based Haptic Rendering with Fast Contact Detection

This paper presents a haptic system which has appropriate frequency to contribute much more real feeling to the users. In this system, haptic rendering is accomplished using streaming point cloud data of Kinect. A preprocessing phase including downsampling and noise reduction is applied to the raw point clouds to prevent huge computation load. In preprocessing phase, it should be concerned that the density of points firmly threats the proxy situation to be in popthrough. Furthermore, the contact is detected by a proxy based algorithm and using a kd-tree search to find contact point among lots of points derived from Kinect. Force can be computed by using a virtual spring coupling between HIP and proxy. The noise is reduced in first phase to reduce the next errors in force calculation. This feature helps us to have a stable system in contact with a dynamic environment including movement and new objects.