In a working environment where robotic procedures are taking place, robots work either in combination with people or not. When the robot parts or its tools which curries, are getting accidentally in touch with human, a collision occurs which is able to cause injuries, accidents or even more damages to the robot. These dangers widow imperative the prevention of such accidents, through methods of detecting collisions, stopping the robot or even react when the robot comes to non-desired contact with humans.
In this work we refer, in first level, all the existing methods of collisions detection between robot and humans, during a literature review. Then, the detection method of dynamic model of the robot is chosen, according to which we form an algorithm which recognizes the conflicts experienced by the robot. The next step is to perform experiments of the choosen method, motioning only one joint, either in a location where gravity is neglected or not. In addition, we export the results of the experiments, summarizing them in a summary table, where we focus on how fast our method recognizes the conflicts and how sensitive is our system to give accurate value of the collision volume. Finally, we compare our method with the model which robot uses in order to identify the external torques, through which is estimated the intensity of the conflict and we complete with the outcome of the conclusions.