Abstract
The original definition of the problem of optimal node visitation (ONV) in acyclic stochastic digraphs concerns the identification of a routing policy that will enable the visitation of each leaf node a requested number of times, while minimizing the expected number of the graph traversals. The original work of Bountourelis and Reveliotis (2006) formulated this problem as a Stochastic Shortest Path (SSP) problem, and since the state space of this SSP formulation is exponentially sized with respect to the number of the target nodes, it also proposed a suboptimal policy that is computationally tractable and asymptotically optimal. This paper extends the results of Bountourelis and Reveliotis (2006) to the cases where (i) the tokens traversing the graph can “split” during certain transitions to a number of (sub-)tokens, allowing, thus, the satisfaction of many visitation requirements during a single graph traversal, and (ii) there are additional visitation requirements attached to the internal graph nodes, which, however, can be served only when the visitation requirements of their successors have been fully met. In addition, the presented set of results establishes stronger convergence properties for the proposed suboptimal policies, and it provides a formal complexity analysis of the considered ONV formulations. From a practical standpoint, the extension of the original results performed in this paper enables their effective usage in the application domains that motivated the ONV problem, in the first place.
Similar content being viewed by others
Notes
That depends on the graph structure and the performance parameters ε and δ.
This can happen due to the stochastic nature of the task transitions.
We remind the reader that a multi-set defined on a set X is essentially a vector ν of dimensionality |X| and with elements belonging to \(\mathbf{Z}_{0}^+\), the set of non-negative integers. Each component ν(i) of vector ν corresponds to one of the elements of X and its value indicates how many replicates of this element are included in the multi-set represented by ν.
We remind the reader that in the QSAT problem we are given a quantified boolean formula with alternating quantifiers, \(\exists x_1 \forall x_2 \exists x_3\hdots \forall x_n, \phi(x_1,\hdots,x_n)\) and we seek to determine whether this formula is satisfiable, that is, whether there is a truth value for x 1 such that for all truth values of x 2, etc. there is a truth value of x n , such that ϕ comes out true.
The gist of this argument is as follows: Consider the “dual LP” (Bertsekas 2005) of the MDP that corresponds to the SSP formulation of the considered ONV problem. Then, any feasible solution of this formulation admits a flow interpretation on the state space of the ONV problem (Bertsekas 2005). Furthermore, the aggregation of this flow, that traverses the state space of the ONV problem, across the arcs of the underlying state transition diagram that correspond to the same transitions in the problem defining graph \({\cal G}\), will provide another flow that constitutes a feasible solution to the relaxing LP. In addition, the original and the induced flows result in the same objective values for their corresponding formulations. But then, it is clear that the relaxing LP is indeed a relaxation of the original ONV formulation and Eq. 13 follows from this result.
We remind the reader that \(f(n)=O(g(n)) \Rightarrow \exists c, n_0\) s.t. 0 ≤ f(n) ≤ c·g(n), ∀ n ≥ n 0.
This bias is established during the policy construction by the structure of the employed optimal solution χ * of the relaxing LP.
Obviously, for nodes x ∈ X L, Succ(x) = ∅ and the condition in the “if” statement of item (3) is immediately satisfied.
Confining this analysis to the set of deterministic policies is enabled by the relevant MDP/SSP theory that guarantees the existence of a deterministic optimal policy.
And not for V *, which was the case with the fluid relaxation of the ONV problem presented in Section 2.
References
Bertsekas DP (1999) Nonlinear programming (2nd ed). Athena Scientific, Belmont
Bertsekas DP (2005) Dynamic programming and optimal control (3rd ed). Athena Scientific, Belmont
Bertsimas D, Gamarnik D (1999) Asymptotically optimal algorithms for job shop scheduling and packet switching. J Algorithms 33:296–318
Bertsimas D, Sethuraman J (2002) From fluid relaxations to practical algorithms for job shop scheduling: the makespan objective. Math Program 92:61–102
Bertsimas D, Tsitsiklis JN (1997) Introduction to linear optimization. Athena Scientific, Belmont
Billingsley P (1968) Convergence of probability measures. Wiley, New York
Bountourelis T, Reveliotis S (2006) Optimal node visitation in acyclic stochastic digraphs. In: Proceedings the 8th intl workshop on discete event systems (WODES’06), IFAC, Ann Arbor, July 2006, pp 358–365
Bountourelis T, Reveliotis S (2007) Rollout policies for the problem of optimal node visitation in acyclic stochastic digraphs. In: European control conference 2007. IEEE, Piscataway, pp 2456–2463
Bountourelis T, Reveliotis SA (2008) Customized learning algorithms for episodic tasks with acyclic state spaces. School of Industrial & Systems Eng., Georgia Tech, Tech Rep
Chen H, Yao DD (2001) Fundamentals of queueing networks: performance, asymptotics, and optimization. Springer, New York
Dai JG (1999) Stability of fluid and stochastic processing networks. Center for Mathematical Physics and Stochastics, University of Aarhus, Denmark, Tech Rep ISSN 1398-7957
Gut A (1974) On the moments and limit distibutions of some first passage times. Ann Probab 2(2):277–308
Meyn S (2008) Control techniques for complex networks. Cambridge University Press, Cambridge
Niño–Mora J (2001) Stochastic scheduling. In: Floudas CA, Pardalos PM (eds) Encyclopedia of optimization. Kluwer, Dordrecht, pp 367–372
Papadimitriou CH (1985) Games against nature. J Comput Syst Sci 31:288–301
Pinedo M (2002) Scheduling: theory, algorithms and systems (2nd ed). Prentice Hall, Upper Saddle River
Reveliotis SA (2007) Uncertainty management in optimal disassembly planning through learning-based strategies. IIE Trans 39:645–658
Reveliotis SA, Bountourelis T (2007) Efficient PAC learning for episodic tasks with acyclic state spaces. J Discrete Event Syst Theory Appl 17:307–327
Reveliotis SA, Bountourelis T (2008) Optimal flow control in acyclic networks with uncontrollable routings and precedence constraints. School of Industrial & Systems Eng., Georgia Tech (under review in IEEE Trans Automat Contr), Tech Rep
Ross SM (1996) Stochastic processes. Wiley, New York
Acknowledgement
This work was partially supported by NSF grants DMI-MES-0318657 and CMMI-0619978.
Author information
Authors and Affiliations
Corresponding author
Additional information
An abridged version of this manuscript was presented at WODES’08.
Appendix: Proof of Lemma 1
Appendix: Proof of Lemma 1
Let ψ′ n = min {k: S k > n·c}. Then ψ′ n is a stopping time and, from Lemma 2.3 of Gut (1974), we have that
where C(r,E[X r]) is a constant depending only on r and E[X r]. Equation 63 further implies that
From Eq. 64 and Theorem 2.3 of Gut (1974), we get
which implies the uniform integrability of \(\left\{n^{-r/2}\cdot\left(\sum_{i=1}^{\psi'_n}(X_i-\mu)\right)^r, n\geq 1\right\}\) (Billingsley 1968).
By the definition of the renewal process ψ′ n ,
which further implies that
Equation 67 combined with the triangle inequality and the fact that
also imply that
and based on the inequality (a + b)r ≤ 2r − 1·(|a|r + |b|r), a,b ∈ R, we finally get
Hence, the uniform integrability of \(\{n^{-r/2}\cdot(\sum_{i=1}^{\psi'_n}(X_i-\mu))^r, n\geq 1\}\) and Eq. 70 imply the uniform integrability of \(\{n^{-r/2}\cdot(n\cdot c- \mu\cdot\psi'_n)^r, n\geq 1\}\). Since ψ′ n = ψ n + 1 we have that
which gives
and implies the uniform integrability of \(\{n^{-r/2}\cdot(n\cdot c-\mu\cdot\psi_n)^r,\ n\geq 1\}\).
Rights and permissions
About this article
Cite this article
Bountourelis, T., Reveliotis, S. Optimal Node Visitation in Acyclic Stochastic Digraphs with Multi-threaded Traversals and Internal Visitation Requirements. Discrete Event Dyn Syst 19, 347–376 (2009). https://doi.org/10.1007/s10626-009-0065-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10626-009-0065-8