Automated Planning Applied In Robot Trajectory Control System

Abstract

Abstract A robot is a group of electromechanical or biomechanical devices, capable to perform an autonomous or preprogrammed work. Robots are commonly used to execute tasks in location with poor illumination, or hazardous conditions for humans and one of the most common application of robots is in the production and transfer lines. Historically the robot programming is based on a set of predefined sequential actions. In this case, devices and sensors are connected to the robot controller and the programming of the jobs or activities is held directly by its internal software. This type of programming is usually fixed and has a little or no flexibility concerning to handling, requiring the equipment reprogramming for new trajectories. In this context there are high cost as result of the labor required and production loss due to setup time for robot training. On the other hand, several studies have been showing artificial intelligence techniques like neural nets, genetic algorithms, automated planning as enhancement proposals for practical systems. Specifically, automated planners emerged in 1971 with the STRIPS or “Stanford Research Institute Problem Solver”, the first automatic solver problems. Automated planning model makes use of search algorithms defined in the model to find a path that starts from the set of inputs and achieve the desired state. In 2008, itSIMPLE was developed as a knowledge engineering tool used for modeling planning domains to several automatic planners. The proposal of this paper is the integration of automated planners with robots control systems through itSIMPLE modeling for trajectory planning. Such integration is accomplished using a device made especially for this project, which is responsible for translating the information served from sensors into states, which will be used by automatic planners. The generated plan is converted into a list of instructions (jobs) that is sent to the robot. Finally, a comparison will be made with the traditional method of programming and robot's training.