ASIC Design and Implementation of the Real-Time Collision Detection for Machine Tool Automation

Abstract

Collision detection of two objects plays an essential role for the machine tool automation. Although the collision detection of two objects has been studied in applications like virtual reality, the collision detection for the machine tool requires high precision to avoid overcut damage to the high-cost machine tools. The current collision detection for machine tools is under the soft-ware based computer numerical control (CNC), the low computation capability of which refrains the CNC based approach from real-time collision detection. In this paper, we consider the design of application specific integrated circuit (ASIC) to enhance the collision detection for machine tool automation. Because the bounded objects are represented by meshed triangles, we consider the separating axis theorem (SAT) based detection algorithm. Furthermore, by considering high precision required by machine tool applications, the proposed algorithm includes collision detection of either non-coplanar or coplanar triangles. Following the collision detection algorithm, we design hardware architecture with parallel processing to provide higher throughput rate over the architecture reported in our conference paper. The VLSI implementation results under the TSMC TN40G (45nm) CMOS technology reveal that our architecture requires 1,212K gates and provides detection throughput 38.46M per second for collision detection of two triangles, while operating at 500 MHz. For two objects represented by 400 and 400 meshed triangles, respectively, our hardware architecture can provide collision detection in 0.96 ms, which is smaller than the 1 ms required for real-time processing of collision detection of two objects.