Deadlock avoidance policies for automated manufacturing cells

Abstract

This chapter considers the problem of deadlock avoidance in flexibly automated manufacturing systems, one of the most prevalent supervisory control problems that challenges the effective deployment of these environments. The problem is addressed through the modeling abstraction of the (sequential) resource allocation system (RAS), and the pursued analysis uses concepts and results from the formal modeling framework of finite state automata (FSA). A notion of optimality is defined through the notion of maximal permissiveness, but the computation of the optimal DAP is shown to be NP-Hard. Hence, the last part of the chapter discusses some approaches that have been developed by the relevant research community in its effort to deal with this negative complexity result. IntroductionContemporary manufacturing is characterized by (i) an ever increasing emphasis on automation and by (ii) a quest for broader product lines and more customized product offerings [22].Indeed, since the beginning of the industrial revolution, automation has been at the core of the mass-production concept and practices, enabling activities and tasks that frequently transcend the human element in terms of capability, consistency, durability and speed [21]. More recently, automation has also been advocated as a key enabler for controlling the overall production cost, especially in economies where labor cost is particularly high. Hence, there is an increasing tendency to confine the presence of the human element on the production shop-floor to supervisory and maintenance tasks only, while the actual production activity is taken over by "intelligent" hardware, materialized in the form of numerically controlled (NC) machines, CHAPTER 6. 2 robots, and other material handling equipment. Clearly, enabling the reliable deployment of such an extensive mode of automation is a pretty challenging task. Yet, in the recent years, things have been further complicated, and the aforementioned challenges have been further exacerbated, by the second manufacturing trend mentioned above, i.e., the emphasis that is placed by the global markets upon choice and customization. This trend implies a need to support the production of a broader set of items, in smaller batch sizes, and with shorter life spans. Such demand and production patterns negate the economies of scale and the efficiencies of the dedicated production lines that have been sought by traditional manufacturing, and (re-)places the emphasis on concepts like resource sharing, flexibility, reconfigurability and concurrency [18]. But the current reality has also shown that the effective support of these concepts on the manufacturing shop-floor is highly non-trivial, and it adds an entirely new layer of complexity and challenges to the aforementioned endeavors towards the automated operation of production systems. The manufacturing world has become increasingly conscious of the fact that the successful design and deployment of flexibly automated production systems necessitates the development of a (ma...

Necessary and Sufficient Conditions for Deadlocks in Flexible Manufacturing Systems Based on a Digraph Model

“…Another formalism is to describe the manufacturing system using graphs [3,5,[8][9][10][11]. In this approach the vertices represent resources and the edges represent product part flows between resources.…”

“…Barkaoui and Abdallah [2], in addition to control place augmentation, used a one step look-ahead controller, which cannot avoid impending deadlocks that are more than one step away. Lawley and Reveliotis [10] identified special classes of systems that present no impending deadlocks.…”

Policy mixtures: a novel approach for enhancing the operational flexibility of resource allocation systems with alternate routings

“…A key requirement for the behavioral correctness of these systems is that they remain live, i.e., that they effectively avoid deadlocking situations where a set of executing processes is permanently blocked due to the fact that each process in this set requires for its further advancement the allocation of some resource(s) held by another process in this set. In [1], it was shown that the problem of deadlock avoidance in sequential RAS translates to the synthesis of a supervisor that will constrain the system behavior in a strongly connected component of its underlying state space, and also contains the system initial empty state. Given the current emphasis on increased resource utilization and operational flexibility, the objective is to obtain the maximally permissive liveness-enforcing supervisor (LES) that will provide the maximum behavioral latitude.…”

“…However, the implementation of the maximally permissive LES is a computationally intractable problem for most real-life applications [2]. Hence, a large body of the past results has been centered around the development of suboptimal but polynomially computable LESs that seek to assure the system liveness while maintaining considerable flexibility, through pertinent exploitation of the available information on the system structure and dynamics [1], [3]- [6].…”

“…In particular, algebraic LESs have been proposed in the past as mathematically elegant and computationally efficient solutions to the problem of LES synthesis for various RAS classes, able to provide a viable tradeoff between computational tractability and operational efficiency [1], [5]- [8]. A formal definition of this class of policies is provided in the next section, but at this point we notice that their algebraic character stems from the fact that they attain the system liveness by constraining the maximum number of jobs that can simultaneously execute various subsets of the RAS process stages.…”