A Method for Hardware Flexibility Evaluation of Laboratory Automation Architectures

---

By - Peyman Najmabadi, PhD, Andrew A. Goldenberg, PhD, Andrew Emili PhD

Abstract
Flexibility is considered one of the major attributes of laboratory automation systems, particularly for the sizeable community of small to medium-size biotechnology research laboratories.

This manuscript briefly explains a method for evaluating the hardware flexibility of laboratory automation architectures for sample preparation protocols. This method provides a quantitative measurement tool for lab automation systems' hardware flexibility without getting into detailed design. As a result, it can be effectively used to select the best flexible automation architecture for a laboratory based on the lab's flexibility requirements. Furthermore, it can be used to identify system-level flexibility shortcomings of conventional laboratory automation architectures and propose new concepts with improved flexibility levels.

Hardware Flexibility of Laboratory
Automation Architectures
Flexibility of laboratory automation systems can be generally defined as the system's ability to be adapted to a new laboratory protocol with the minimum cost and downtime of the existing system. This article focuses on automation architectures for sample preparation protocols. These protocols usually consist of several liquid-handling and sample-processing steps. Automation systems for such protocols should integrate varied laboratory processing instruments in a workcell equipped with automated means for sample manipulation operations (liquid handlings and transportation of laboratory containers). Furthermore, total flexibility of a laboratory automation system is affected by both the system's hardware and software. In this article, hardware flexibility or system components' adaptability to protocol variations has been considered.

Hardware flexibility modelling of laboratory automation systems can be performed by introducing several parametric measures. These parameters should be chosen in such a way that each parameter explains one aspect of hardware flexibility. With regard to system-level hardware flexibility modelling of sample manipulation operations of laboratory automation systems, three major parameters of functional, structural and throughput have been proposed.1,2 It should be noted that the mentioned list of flexibility parameters is not the comprehensive list. It does, however, show major hardware flexibility characteristics related to sample manipulation operations and system layout. In the following, the aforementioned parameters are defined, followed by explanation of a quantitative measurement method for these parameters.

Hardware Flexibility Measurement
In order to measure hardware flexibility parameters defined in the previous section, a method in the realm of the Axiomatic Theory for Design has been proposed. This theory provides a powerful tool for the synthesis and evaluation of conceptual designs. It consists of two axioms, several corollaries and theorems that help creation and analysis of conceptual designs. A method similar to Axiom 2 of the Axiomatic Theory is proposed for the measurement and evaluation of hardware flexibility parameters. Before explaining this method, the rationale of Axiom 2 is described briefly below.

Axiom 2 has been proposed for the evaluation of conceptual designs and selection of the optimum design. It is stated as "among designs that are consistent with Axiom 1 (the axiom which validates that all design requirements have been satisfied appropriately through a proposed conceptual design), one that has the highest probability of success is the best design." It can be also stated as "the design with the minimum information content has to be selected," as the information content is proportional to the inverse of probability of success.3 In order to implement Axiom 2, the probability of satisfying design requirements through a proposed conceptual design should be calculated. This starts with defining the probability of success functions for each of the design requirements. Considering pi as the probability of success function for the ith design requirement, the information content of a conceptual design in the case of having n design requirements can be calculated as seen in Equation 1 (see below).

The information content formulation is in a logarithmic form in order to express the sum of information contents of various requirement parameters of a design. The logarithm can be in base two or natural logarithm, which makes the unit of information content as bits or nats, respectively.

A similar approach has been adopted for the measurement of hardware flexibility parameters. The strategy is to define the probability of success functions for flexibility parameters. These functions basically show how flexible an automation system concept is with respect to each of the flexibility parameters. Then, similar to Axiom 2, we define flexibility information contents by taking the logarithm of flexibility probability functions. The logarithm operation has been employed to express the sum of information on different flexibility parameters. Similar to Axiom 2 of Axiomatic Theory, we can state that a more flexible laboratory automation system would have higher flexibility probability of success or lower flexibility information content.1,4 We can use Equation 1 to calculate sum of flexibility information content on different flexibility parameters. The main advantage of using this method as a measure for hardware flexibility of laboratory automation systems is that it provides a tool for flexibility evaluation of the general scheme of a laboratory automation design, before getting into detailed design.

Probability function for each of the flexibility parameters should be defined in such a way that its value becomes "1" when an automation system is fully flexible with regard to the corresponding flexibility parameter (probability of success is 100%). In this case, the information content is calculated as the lowest possible value, or zero. Probability functions for two of the flexibility parameters indicated before (transportation and throughput flexibilities) are provided in the following as examples to show how the aforementioned method can be used for flexibility evaluation of laboratory automation systems. A thorough investigation of this method for the measurement of hardware flexibility parameters is provided in Reference 1.

Hardware Flexibility Probability Function
A. Probability Function of Transportation Flexibility (Ff,t )
Based on the definition provided for the transportation flexibility parameter, probability of success function can be defined as Equation 2.
The denominator in Equation 2 is the number of entities in a group consisting of different required types of containers based on the requirements of working protocols in a laboratory. Based on the transportation mechanism employed in a laboratory automation system, the system is capable of transporting some or all types of containers. The number of common entities between this group and the group of denominator is the numerator indicated in Equation 2, which shows how flexible an automation system is in terms of transportation according to the lab's requirements. Information content of transportation flexibility can be also calculated by Equation 3.

B. Probability Function of Throughput
Flexibility (Fv )
Throughput flexibility pertains to the capability of a system to perform parallel processing. As a result, this parameter is related directly to the workspace expandability of an automation system. Probability of success for throughput flexibility (Fv) can be defined as Equation 4.

The numerator of Equation 4 is the maximum possible number of parallelly employed instruments in a system for a protocol or a range of protocols based on the workspace expandability specification of an automation system. The denominator of Equation 4 is the required number of parallel instruments based on the required throughput of a laboratory for its working batch protocols. Information content of the throughput flexibility can be calculated by Equation 5 (see below).

Similarly, probability functions for other hardware flexibility parameters can be defined and flexibility information contents can de calculated. As mentioned before, total flexibility information content is the summation of all flexibility information content. As the flexibility information content decreases, the flexibility of an automation system increases. In the following, transportation and throughput flexibilities of two of the conventional laboratory automation architectures are evaluated using the formulations indicated in this section.

Conventional Laboratory Automation Architectures
Laboratory automation systems can be categorized based on different attributes such as means of transportation, application, flexibility and cost. With respect to the sample transportation or manipulation mechanism, they can be classified into two major groups: (i) robotic-based and (ii) track-based.5 In the robotic-based approach, a robotic manipulator is used for transportation of containers between different laboratory instruments.

In this approach, a robotic arm (cylindrical, articulated, Cartesian, etc.) shuttles single or several containers between different laboratory instruments. However, in the track-based approach, a transportation mechanism (such as conveyor) is used instead of the robotic arms for transportation of containers between different processing instruments. These are the two basic laboratory automation approaches, while a combination of them in different configurations also exists.

With regard to transportation flexibility evaluation (Equation 2 and Equation 3), different types of containers that are used in biotechnology protocols should be considered. Containers can be generally classified into three major groups: tubes, plates and rack of tubes. Therefore, the denominator in Equation 2, which stands for the required types of containers for transportation in an automation system is “3.” However, the numerator is based on the transportation mechanism of an automation system.

A track-based transportation system is usually capable of transporting one type of container (e.g. plates or single tubes). Therefore, the numerator of Equation 2 for the track-based automation approach is “1.” As a result, probability function for this flexibility parameter in the case of track-based approach can be calculated as (1/3). The fact that we cannot reach to the maximum probability of success value, which is “1,” shows that the conventional track-based approach can not provide the maximum transportation flexibility level. Information content of transportation flexibility of the track-based approach can be also calculated based on Equation 3 as Equation 6.

Nevertheless, in the case of the robotic-based approach, transportation is performed by a robotic manipulator. Using a simple finger-type gripper with appropriate size, it is possible to transport all of the aforementioned types of containers. Therefore, the probability of success (Equation 2) in this case is “1” and as a result the information content for transportation flexibility in the case of the robotic-based approach is the minimum possible value or zero as shown by Equation 7.

Equation 7 shows that it is possible to reach to the maximum flexibility level in terms of transportation flexibility through the robotic-based automation approach. If an automation system includes both track-based and robotic-based approaches, their transportation flexibility can be measured similarly through Equation 2 and Equation 3 based on the total capability of the employed transportation system.

With regard to throughput flexibility, in the case of the track-based approach, there is no limitation in adding new laboratory instruments around the track to perform parallel processing because the track can be extended to access all instruments. Therefore, the highest level of flexibility in terms of throughput flexibility can be achieved through the track-based approach or its probability function is “1” ((Pv)track-based=1).
However, throughput flexibility of the robotic-based approach is limited because of the limited workspace size of robotic manipulators. In other words, it is not always possible to reach to the required throughput by parallel processing in the case of robotic-based systems.

For example, assume that in a robotic-based laboratory automation system there is enough space for incorporating “n” laboratory instruments in parallel. This system can provide a fully flexible system in terms of throughput flexibility for laboratories that do not need incorporating more than “n” instruments. For instance, assume that based on the throughput requirements of a laboratory, up to “2n” processing instruments might be needed. In this case, throughput flexibility probability function of the aforementioned robotic-based system can be calculated as (1/2) (Equation 8). This value shows how flexible the automation system is in terms of throughput requirement of the laboratory. It can be also used for flexibility comparison of different automation systems based on the laboratory throughput requirement by Equation 8.

Conclusion
A method for quantitative evaluation of hardware flexibility of laboratory automation architectures was introduced. This method is based on calculation of a laboratory automation system’s flexibility information content. As flexibility information content increases, flexibility of an automation system decreases. This method can be used to investigate appropriateness of an automation system in terms of flexibility for a laboratory. It can also provide a guideline for selecting the best flexible automation system. The selection is performed based on the lowest flexibility information content criterion. Ideally, an automation system with zero information content for all flexibility parameters is the one that provides the highest flexibility level in terms of all parameters for a laboratory. Furthermore, the aforementioned measurement method can be used to recognize hardware flexibility shortcomings of conventional laboratory automation architectures. As a result, guidelines for development of new automaton concepts that address specific flexibility shortcomings can be identified. This method has been applied for identification of flexibility shortcomings of conventional track-based and robotic-based automation approaches. New concepts to address their flexibility shortcomings can be seen in an elaborated version of this manuscript, which will appear in the August 2006 volume of Journal of the Association for Laboratory Automation.

REFERENCES
1. Najmabadi, P., A.A. Goldenberg, and A. Emili. “Hardware flexibility of laboratory automation systems: analysis and new flexible automation architectures.” Journal of the Association for Laboratory Automation., Accepted for publication in August 2006 volume.

2. Najmabadi, P., A.A. Goldenberg, and A. Emili. “A system concept for flexible implementation of robotic-based automation in biotechnology research laboratories.” Proc. IEEE International Conference on robotics and Automation, Orlando, May 2006, 4297-99, to be published.

3. Suh, N.P., E. Ralph, and F. Eloise. “Axiomatic design theory for systems.” Research in Engineering Design, 10.4, (1998): 189-209.

4. Najmabadi, P., A.A. Goldenberg, and A. Emili. “New flexible laboratory automation system concepts for biotechnology research laboratories.” Presented at LabAutomation 2006 Conference organized by ALA (Association for Laboratory Automation), Palm Spring, CA, January 21-25, 2006.

5. Najmabadi, P., A.A. Goldenberg, and A. Emili. “A scalable robotic-based laboratory automation system for medium-sized biotechnology laboratories.” Proc. IEEE
International Conference on Automation Science and Engineering, Edmonton, Canada, (2005): 166-171.


Peyman Najmabadi, PhD is a resarch assistant at the Robotics and Automation Lab in the department of mechanical and industrial engineering, University of Toronto (U of T) (Toronto, ON). Andrew A. Goldenberg, PhD is a professor in U of T’s department of mechanical &
industrial engineering, and Andrew Emili, PhD is an associate professor in the Program in Proteomics and
Bioinformatics at U of T’s Banting and Best Department of Medical Research.