A Fault detection and Isolation (FDI) Algorithm for a virtual simulator of an industrial robot is developed [17] in order to make online supervision for possible collisions detection without using any vision feedback surveillance system. The real and virtual systems are operating in parallel, while the algorithm developed runs in real-time. The approach can alert the operator to predict dangerous and accidental situations before it occurs. The system is able to detect collision for industrial robot but its avoidance needs further autonomy to remove human    intervention.

Vision based path planning is by far the most efficient approach present for path planning in industries because of its many advantages such as minimum cost, easy installation, capture the real map of the scene and collision detection before the actual contact, etc. Therefore, vision based machine surveillance, monitor- ing and guidance is one of the important approach applied in various industries. One of the important study for human–robot collaborative environment using vision-guided active collision avoidance system [18] focuses on better flexibility and productivity. In multi-axis machines, there exist many collision detection approaches such as C-space approach [19,20] and visibility data approach [21,22] but collision avoidance is still an area which needs further consideration. Our previous vision based approaches [9–11] gives promising results for safe tool-path generation in multi-axis production but a high speed tool-path needs a rapid decision with collision avoidance from a safe distance for non- functional trajectories. An intelligent rapid decision making method is required to deal with off-line and real-time collisions in a machining environment where objects are detected automati- cally using vision systems (2D/3D). A vision based algorithm presented previously [10,11] can detect and avoid collisions intelligently but 2D camera installed in a scene may not take into account depth information and occlusion in the scene, which may lead to false collision detection. Multiple cameras [9] may solve this problem to a certain extent but again it will produce good results  linked  with  high  equipment  cost  and  increase  in        the

computation time.

Laser scanners are also widely used to detect the environment especially for mobile robots. For instance, a system called 3D SLAM [23], for modeling the environment and collision detection and avoidance is based on projection techniques which works for short range. Laser scanners usually requires multiple projections of lines or  patterns  to  get  data  from  the  scene,  therefore  they  are   not

Fig. 1. Universal milling machine (Deckel Maho DMU 50).

suitable for dynamic scenarios especially for real-time CNC manufacturing and rapid traversal trajectories. A time-of-flight camera is known to be the best choice [24] for dynamic scenarios in a  3D environment.

A 3D Modulated Light Intensity (MLI) sensor camera also called as Time-of-Flight (ToF) camera [25] (IEE industry) is installed to observe the universal milling machine shown in Fig. 1. The ToF cameras are known to be cost-efficient, 3D sensing systems, which are capable of providing full depth information from the scene at a high frame rate regardless of the illumination conditions and with no texture dependencies. The ToF sensors are based on the principle of measuring the time the light emitted (nearly infrared) by the active sensor, needs to reach the surface being observed and receiving its reflection to calculate the depth    [26].

This 3D ToF sensor provides depth information regarding the scene that will be used by the method presented  to detect collisions with the static obstacle (initial application) in the scene as shown in Fig. 2. The obstacle shown is an imaginary polymer cube, which represent an unknown and un-programed obstacle (machine element/fixture, etc.) present in the manufacturing scenario. Once a potential collision is detected, the approach presented in the next section, will provide two different choices for safe tool-paths to the machine. One of the best trajectory between two possible solutions generated automatically, is selected according to certain defined criteria for avoiding tool collision intelligently  without  any  human intervention. 铣削和钻削数控机床英文文献和中文翻译(2):http://www.chuibin.com/fanyi/lunwen_206129.html