Users of modern packaging lines expect a lot from their installations. Typically, products of varying size and shape are presented for packaging at random intervals and in random orientations. In principle, robot technology offers the flexibility needed to cope with these variations, but it can't do so unless it has some way of accurately determining product size, position and orientation.
The central ingredient of the most versatile solutions is a machine vision system but, in itself, such a system is not a complete solution. Almost any modern machine vision system from a reputable manufacturer will be capable of producing good quality data, but the real issues are how that data is used, and how easy the system is to work with.
It's important to bear in mind that machine vision system manufacturers produce equipment that can be used in a wide range of applications; it is not specifically optimised for use with robots. This means that although the equipment offers all of the facilities - such as geometric pattern matching and blob analysis - needed for robot control, it will often be unnecessarily difficult for the end user of the packaging machine to access and use these facilities.
The first problem that needs to be addressed is communication between the vision system and the robot controller. This is, of course, an issue for the supplier of the packaging machine, but it's important to users in that fast and dependable communication is essential for achieving high levels of productivity. Users expect today's packaging lines to be very productive, with average throughput rates of several hundred packs per minute being quite typical. This high speed of operation places considerable demands on the robot controller - powerful processors are needed, coupled with software which implements fast but stable control loops for the robot motion. However, all of this is of little value if the communication link with the vision system is slow or unreliable.
Even the best software and best communications links are of little value unless end-users can take full advantage of their potential. Until recently, this was more difficult than it sounds - both vision systems and robots had a deserved reputation for being hard to set up. Fortunately, developments in user interface technology mean that this is no longer the case. The best robotic systems have graphical programming systems which are based on familiar Windows technology and which are very intuitive in use. The days of developing manually coded programs for each application and even each product are over.
In addition, leading suppliers of robotic packaging machines provide their own interface for the vision system. Instead of the user having to grapple with the full range of functions offered by the vision system manufacturer, many of which are irrelevant in robot applications, they are presented with selected functions which have been optimised for use with robots. Essential operations, such as system calibration and object identification, which have, in the past been tedious and time consuming, can be automated. Users can also be provided with a library of functions - such as checking that every position in a tray of products is properly filled - instead of having to programme these functions from scratch. The result is big savings in machine set up time.
Tight integration between the robot controller and the vision system can also provide the basis for other benefits, such as automatic sharing and balancing of the workload between multiple robots used in a packaging machine. This enables throughput and operating efficiency to be maximised.
This article has, until now, concentrated on the aspects of machine vision systems that relate directly to motion control, but they can also provide other functions, such as automatic inspection. A vision system can inspect the product itself, detecting all kinds of visual defects such as incorrect shape and poor colour. The system can also inspect the packaging, revealing problems like missing or illegible date stamps, missing labels and incorrectly filled packs. Best of all, in most cases, inspection can be accomplished with the same vision system that is being used for control purposes. The incremental cost of this additional functionality is, therefore, minimal but the savings it delivers by doing away with manual inspection can be very substantial.
In packaging applications, vision systems should be considered as an integral part of every robot-based installation. But the performance of the vision system itself is a relatively minor concern when evaluating the installation's overall capabilities.
Frank-Peter Kirgis is with ABB Robotics