Open Real-Time Robot Controller Framework

Abstract

Abstract The challenge with advanced robot control in manufacturing is two-fold, regarding industrial robot controllers:1) General real-time control from external entities are not supported; and only for special cases of applicationscenarios, limited real-time extensions to the controller can be purchased. 2) The robot controller application-platforms are robot centric; leaving an external application to battle with achieving the desired behaviour.Based on free and open software resources, experiments have been performed with three industrial robot con-trollers, and measurements of response times and tracking delay from external control are presented. Alsopresented is the design of a motion control framework, demonstrating external integration of force feedback andvisual servoing. Keywords: Manufacturing system; Real-time control; Robot motion control; Robot sensor-servoing 1 INTRODUCTION The past couple of decades have seen an ever increasing de-mand for flexibility and adaptability in manufacturing automa-tion. Regarding the near future of manufacturing in westerncountries, a quite probable scenario is that manufacturing ofsimple goods with no or little variation, will be almost non-existing. The manufacturing industry that will remain in thispart of the world will have emphasis on a high degree of cus-tomization, almost to the level of having no product catalogue.One of the drivers for this effect is the market demands, re-questing customization and personalization, simply due to thepossibility [1]. This is a stimulative and additive effect, changingexisting manufacturing companies and shaping new ones. An-other driver for this is the outsourcing or relocation of uncompli-cated large-series production to low-cost countries. This lattereffect is a subtractive and inhibitory effect in the sense that it re-moves manufacturing companies that do, or can, not change,and prevents establishment of new manufacturing companiesthat only manufacture simple goods.These two effects are, of course, but two among a whole rangeof other effects, and can not encompass the plethora of aspectsand types of manufacturing. However, in the general subjectof manufacturing research, they are the predominant effectsdiscussed regarding automation.Automation is well in the process of taking over shop-floor levelactivities, like processing, handling, and transportation. Fac-tory level activity, like orchestration, real-time (re-)scheduling,and online logistics management, is under development to beautomated, hence closing the gap between Enterprise Re-source Planning (ERP) systems and shop-floor control.MotivationRobot manipulators are used to meet the requirements ofagility, reachability, flexibility, adaptability, dexterity, etc., inmanufacturing systems. The most flexible kind of industrial ma-nipulator is the (serially linked) articulate robot, and it mostlyhas 6 degrees of freedom (DOF). A challenge with such mech-anisms is, that the mapping between the actuator and opera-tional spaces are highly non-linear. This is why an advancedmotion controller is always found associated with such a robot.Historically, there has not been an overall application controlat the factory or shop-floor levels, so robot-centric applicationcontrollers were implemented co-located with the motion con-trollers in the robot controllers. These have evolved to quite ad-vanced platforms, but typically remain closed and proprietary,shielding off the underlying servo controller from the applicationprogrammer. Hence, the native application controllers are wellsuited for the use cases that were part of the platform devel-opers’ design criteria, but virtually excludes or hinders all otheruses and application scenarios.To render the robots more general and generic, the world ofrobotics has seen some projects aimed at developing opencontrollers, independent of the robot and controller manufac-turers. Examples of such are the Open Modular Controller, de-veloped by a team led by Jensen [2], and the OROCOS project[3]. Such projects provide an advanced application platform,which is completely open, and thus allowing any applicationscenario within the limits of mechanics, hardware, and real-time communication.Facing the need for application flexibility and factory-wide au-tomated control, the robot controllers are no longer adequateas application platform. Further, the open application-platformcontrollers may also be too complex, since they remain robotcentric. I.e. they still assume that it is within the applicationcontroller of the robot, that the major part of the application isto be implemented.In a distributed, intelligent system for automatic control at fac-tory or shop-floor level, the application platform is in “the sky”;i.e. in the local network in the factory. Such a control system,e.g. a Holonic Manufacturing System [4], will benefit from softreal-time access to motion- or servo-control. Local applicationscenarios, like sensor-servo-based motion control, will needsemi-hard real-time access to the servo-controller [5].Pre-print for 3rd CIRP Conference on Assembly Technologies and Systems, 2010