Skip to main content

Advertisement

Springer Nature Link
Log in

Bayes-adaptive hierarchical MDPs

  • Published:
Applied Intelligence Aims and scope Submit manuscript

Abstract

Reinforcement learning (RL) is an area of machine learning that is concerned with how an agent learns to make decisions sequentially in order to optimize a particular performance measure. For achieving such a goal, the agent has to choose either 1) exploiting previously known knowledge that might end up at local optimality or 2) exploring to gather new knowledge that expects to improve the current performance. Among other RL algorithms, Bayesian model-based RL (BRL) is well-known to be able to trade-off between exploitation and exploration optimally via belief planning, i.e. partially observable Markov decision process (POMDP). However, solving that POMDP often suffers from curse of dimensionality and curse of history. In this paper, we make two major contributions which are: 1) an integration framework of temporal abstraction into BRL that eventually results in a hierarchical POMDP formulation, which can be solved online using a hierarchical sample-based planning solver; 2) a subgoal discovery method for hierarchical BRL that automatically discovers useful macro actions to accelerate learning. In the experiment section, we demonstrate that the proposed approach can scale up to much larger problems. On the other hand, the agent is able to discover useful subgoals for speeding up Bayesian reinforcement learning.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Subscribe and save

Springer+ Basic
$34.99 /Month
  • Get 10 units per month
  • Download Article/Chapter or eBook
  • 1 Unit = 1 Article or 1 Chapter
  • Cancel anytime
Subscribe now

Buy Now

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6

Similar content being viewed by others

Explore related subjects

Discover the latest articles, news and stories from top researchers in related subjects.

References

  1. Abbeel P, Coates A, Quigley M, Ng AY (2006) An application of reinforcement learning to aerobatic helicopter flight

  2. Asmuth J, Littman ML (2011) Learning is planning: near Bayes-optimal reinforcement learning via Monte-Carlo tree search. In: UAI, p 19–26

  3. Atkeson CG (1997) Nonparametric model-based reinforcement learning. In: Advances in Neural Information Processing Systems (NIPS)

  4. Auer P, Cesa-Bianchi N, Fischer P (2002) Finite-time analysis of the multiarmed bandit problem. Mach Learn 47(2-3):235– 256

    Article  MATH  Google Scholar 

  5. Bai H, Hsu D, Lee WS, Vien NA (2010) Monte Carlo value iteration for continuous-state POMDPs. In: Algorithmic foundations of robotics IX, p 175–191

  6. Barto AG, Mahadevan S (2003) Recent advances in hierarchical reinforcement learning. Discrete Event Dynamic Systems 13(4):341–379

    Article  MathSciNet  MATH  Google Scholar 

  7. Baxter J, Tridgell A, Weaver L (2000) Learning to play chess using temporal differences. Mach Learn 40(3):243–263

    Article  MATH  Google Scholar 

  8. Cao F, Ray S, Bottou L (2012) Bayesian hierarchical reinforcement learning. In: 0Bartlett P, Pereira F, Burges C, Weinberger K (eds) Advances in Neural Information Processing Systems (NIPS), pp 73–81

  9. Castro PS, Precup D (2007) Using Linear Programming for Bayesian exploration in Markov decision processes. In: IJCAI, p 2437–2442

  10. Cuayȧhuitl H, Kruijff-Korbayovȧ I, Dethlefs N (2014) Nonstrict hierarchical reinforcement learning for interactive systems and robots. TiiS 4(3):15:1–15, 30

    Google Scholar 

  11. Dearden R, Friedman N, Russell SJ (1998) Bayesian Q-learning. In: AAAI, p 761–768

  12. Dietterich TG (2000) Hierarchical reinforcement learning with the MAXQ value function decomposition. J Artif Intell Res (JAIR) 13:227–303

    MathSciNet  MATH  Google Scholar 

  13. Duff M (2002) Optimal learning: Computational procedures for Bayes-adaptive Markov decision processes. PhD thesis, University of Massassachusetts Amherst

  14. Engel Y, Mannor S, Meir R (2003) Bayes meets Bellman: The Gaussian process approach to temporal difference learning. In: Proceedings of the International Conference on Machine Learning, p 154–161

  15. Engel Y, Mannor S, Meir R (2005) Reinforcement learning with Gaussian processes. In: Proceedings of the International Conference on Machine Learning, p 201–208

  16. Ghavamzadeh M, Engel Y (2006) Bayesian policy gradient algorithms. In: Advances in Neural Information Processing Systems (NIPS), p 457–464

  17. Ghavamzadeh M, Engel Y (2007) Bayesian actor-critic algorithms. In: Proceedings of the International Conference on Machine Learning, p 297–304

  18. Guez A, Silver D, Dayan P (2012) Efficient Bayes-adaptive reinforcement learning using sample-based search. In: Advances in Neural Information Processing Systems (NIPS), p 1034–1042

  19. Hauskrecht M, Meuleau N, Kaelbling LP, Dean T, Boutilier C (1998) Hierarchical solution of Markov decision processes using macro-actions. In: UAI, p 220–229

  20. He R, Brunskill E, Roy N (2010) PUMA: Planning under uncertainty with macro-actions. In: Proceedings of the Association for the Advancement of Artificial Intelligence (AAAI)

  21. Hong J, Prabhu VV (2004) Distributed reinforcement learning control for batch sequencing and sizing in just-in-time manufacturing systems. Appl Intell 20(1):71–87

    Article  Google Scholar 

  22. Iglesias A, Martínez P, Aler R, Fernández F (2009) Learning teaching strategies in an adaptive and intelligent educational system through reinforcement learning. Appl Intell 31(1):89–106

    Article  Google Scholar 

  23. Konidaris G, Barto AG (2009) Skill discovery in continuous reinforcement learning domains using skill chaining. In: Advances in Neural Information Processing Systems 22: 23rd Annual Conference on Neural Information Processing Systems 2009. Proceedings of a meeting held 7-10 December 2009, Vancouver, British Columbia, Canada., p 1015–1023

  24. Lang T, Toussaint M (2010) Planning with noisy probabilistic relational rules. J Artif Intell Res (JAIR) 39:1–49

    MATH  Google Scholar 

  25. Lim ZW, Hsu D, Sun LW (2011) Monte Carlo value iteration with macro-actions. In: Advances in Neural Information Processing Systems (NIPS), p 1287–1295

  26. McGovern A, Barto AG (2001) Automatic discovery of subgoals in reinforcement learning using diverse density. In: Proceedings of the International Conference on Machine Learning, p 361– 368

  27. Mcgovern A, Sutton RS, Fagg AH (1997) Roles of macro-actions in accelerating reinforcement learning. In: In Grace Hopper Celebration of Women in Computing, p 13–18

  28. Pineau J, Thrun S (2001) An integrated approach to hierarchy and abstraction for POMDPs. Tech. rep., Carnegie Mellon University. Robotics Institute

  29. Poupart P, Vlassis NA, Hoey J, Regan K (2006) An analytic solution to discrete Bayesian reinforcement learning. In: Proceedings of the International Conference on Machine Learning, p 697–704

  30. Ross S, Chaib-draa B, Pineau J (2007) Bayes-adaptive POMDPs. In: Advances in Neural Information Processing Systems (NIPS)

  31. Ross S, Pineau J (2008) Model-based bayesian reinforcement learning in large structured domains. In: UAI, pp. 476–483

  32. Samuel AL (1959) Some studies in machine learning using the game of checkers. IBM J Res Dev 3(3):210–229

    Article  MathSciNet  Google Scholar 

  33. Silver D, Veness J (2010) Monte-carlo planning in large POMDPs. In: Advances in Neural Information Processing Systems (NIPS), pp. 2164–2172

  34. Simsek Ö., Barto AG (2004) Using relative novelty to identify useful temporal abstractions in reinforcement learning. In: Proceedings of the International Conference on Machine Learning

  35. Simsek Ö., Barto AG (2008) Skill characterization based on betweenness. In: Advances in Neural Information Processing Systems (NIPS), p 1497–1504

  36. Singh SP, Bertsekas D (1996) Reinforcement learning for dynamic channel allocation in cellular telephone systems. In: Advances in Neural Information Processing Systems (NIPS), p 974–980

  37. Strens MJA (2000) A Bayesian framework for reinforcement learning. In: Proceedings of the International Conference on Machine Learning, p 943–950

  38. Sutton RS, Barto AG (1998) Reinforcement Learning: An Introduction. MIT Press, Cambridge

    Google Scholar 

  39. Sutton RS, Precup D, Singh SP (1999) Between MDPs and semi-MDPs: A framework for temporal abstraction in reinforcement learning. Artif Intell 112(1-2):181–211

    Article  MathSciNet  MATH  Google Scholar 

  40. Tesauro G (1992) Practical issues in temporal difference learning. Mach Learn 8:257–277

    MATH  Google Scholar 

  41. Tesauro G (1994) TD-Gammon, a self-teaching backgammon program, achieves master-level play. Neural Comput 6(2):215–219

    Article  Google Scholar 

  42. Tesauro G (1995) Temporal difference learning and TD-Gammon. Commun ACM 38(3):58–68

    Article  Google Scholar 

  43. Theocharous G, Kaelbling LP (2003) Approximate planning in POMDPs with macro-actions. In: Advances in Neural Information Processing Systems (NIPS)

  44. Vien NA, Chung T (2007) Natural gradient policy for average cost SMDP problem. In: Proceedings of the IEEE International Conference on Tools with Artificial Intelligence, p 11–18

  45. Vien NA, Chung T (2008) Policy gradient semi-Markov decision process. In: Proceedings of the IEEE International Conference on Tools with Artificial Intelligence, p 11–18

  46. Vien NA, Ertel W (2012) Learning via human feedback in continuous state and action spaces. In: AAAI Fall Symposium: Robots Learning Interactively from Human Teachers

  47. Vien NA, Ertel W (2012) Monte carlo tree search for bayesian reinforcement learning. In: 11th International Conference on Machine Learning and Applications, ICMLA, Boca Raton, FL, USA, December 12-15, 2012. Volume 1, p 138–143

  48. Vien NA, Lee S, Chung T (2010) Policy gradient based semi-Markov decision problems: Approximation and estimation errors. IEICE Trans 93-D(2):271–279

    Google Scholar 

  49. Vien NA, Ngo H, Ertel W (2014) Monte Carlo Bayesian hierarchical reinforcement learning. In: Proceedings of the International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS), pp. 1551–1552. International Foundation for Autonomous Agents and Multiagent Systems

  50. Vien NA, Ngo H, Lee S, Ertel W (2014) Approximate planning for bayesian hierarchical reinforcement learning. Appl Intell 41(3):808–819

    Article  Google Scholar 

  51. Vien NA, Toussaint M (2014) Model-based relational RL when object existence is partially observable. In: (ICML 2014)

  52. Vien NA, Toussaint M (2015) Hierarchical Monte-Carlo planning. In: Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence, January 25-30, 2015, Austin, Texas, USA., p 3613–3619

  53. Vien NA, Toussaint M (2015) POMDP manipulation via trajectory optimization. In: Proc. of the Int. Conf. on Intelligent Robots and Systems (IROS 2015)

  54. Vien NA, Toussaint M (2015) Touch based POMDP manipulation via sequential submodular optimization. In: Proc. of the IEEE-RAS Int. Conf. on Humanoid Robots (Humanoids 2015)

  55. Vien NA, Viet NH, Lee S, Chung T (2007) Heuristic search based exploration in reinforcement learning. In: IWANN, p 110–118

  56. Vien NA, Viet NH, Lee S, Chung T (2007) Obstacle avoidance path planning for mobile robot based on ant-q reinforcement learning algorithm. In: ISNN (1), p 704–713

  57. Vien NA, Viet NH, Lee S, Chung T (2009) Policy gradient SMDP for resource allocation and routing in integrated services networks. IEICE Trans 92-B(6):2008–2022

    Article  Google Scholar 

  58. Vien NA, Viet NH, Park H, Lee S, Chung T (2007) Q-learning based univector field navigation method for mobile robots. In: Khaled E (ed) Advances and Innovations in Systems, Computing Sciences and Software Engineering, p 463–468

  59. Vien NA, Yu H, Chung T (2011) Hessian matrix distribution for Bayesian policy gradient reinforcement learning. Inf Sci 181(9):1671–1685

    Article  MathSciNet  MATH  Google Scholar 

  60. Viet NH, Vien NA, Chung T (2008) Policy gradient SMDP for resource allocation and routing in integrated services networks. In: ICNSC, p 1541–1546

  61. Viet NH, Vien NA, Lee S, Chung T (2008) Obstacle avoidance path planning for mobile robot based on multi colony ant algorithm. In: ACHI, p 285–289

  62. Wang T, Lizotte DJ, Bowling MH, Schuurmans D (2005) Bayesian sparse sampling for on-line reward optimization. In: Proceedings of the International Conference on Machine Learning, p 956–963

  63. Wang Y, Won KS, Hsu D, Lee WS (2010) Monte Carlo Bayesian reinforcement learning. In: Proceedings of the International Conference on Machine Learning

  64. White CC (1976) Procedures for the solution of a finite-horizon, partially observed, semi-Markov optimization problem. Oper Res 24(2):348–358

    Article  MathSciNet  MATH  Google Scholar 

  65. Zhang W, Dietterich TG (1995) A reinforcement learning approach to job-shop scheduling. In: International Joint Conferences on Artificial Intelligence, p 1114–1120

Download references

Acknowledgements

The authors are grateful to the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science, and Technology (2014R1A1A2057735) for its tremendous support to this work’s completion.

Author information

Authors and Affiliations

  1. Machine Learning and Robotics Laboratory, University of Stuttgart, Stuttgart, Germany

    Ngo Anh Vien

  2. College of Liberal Arts, Kyung Hee University, 1 Seocheon-dong, Giheung-gu, Yongin-si, Gyeonggi-do, 446-701, Korea

    SeungGwan Lee

  3. Department of Computer Engineering, Kyung Hee University, Seocheon-dong, Giheung-gu, Yongin-si, Gyeonggi-do, 446-701, South Korea

    TaeChoong Chung

Authors
  1. Ngo Anh Vien
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. SeungGwan Lee
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. TaeChoong Chung
    View author publications

    You can also search for this author inPubMed Google Scholar

Corresponding authors

Correspondence to Ngo Anh Vien or TaeChoong Chung.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Vien, N.A., Lee, S. & Chung, T. Bayes-adaptive hierarchical MDPs. Appl Intell 45, 112–126 (2016). https://doi.org/10.1007/s10489-015-0742-2

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10489-015-0742-2

Keywords