How to make reactive planners risk-sensitive (Without Altering Anything)

Probabilistic planners can have various planning objectives: usually they either maximize the probability of goal achievement or minimize the expected execution cost of the plan. Researchers have largely ignored the problem how to incorporate risk-sensitive attitudes into their planning mechanisms. We discuss a risk-sensitive planning approach that is based on utility theory. Our key result is that this approach can, at least for risk-seeking attitudes, be implemented with any reactive planner that maximizes (or satisfices) the probability of goal achievement. First, the risk-sensitive planning problem is transformed into a different planning problem, that is then solved by the planner. The larger the probability of goal achievement of the resulting plan, the better its expected utility is for the original (risk-sensitive) planning problem. This approach extends the functionality of reactive planners that maximize the probability of goal achievement, since it allows one to use them (unchanged) for risk-sensitive planning.

[1]  Simon Parsons,et al.  Do the right thing - studies in limited rationality by Stuart Russell and Eric Wefald, MIT Press, Cambridge, MA, £24.75, ISBN 0-262-18144-4 , 1994, The Knowledge Engineering Review.

[2]  Jon Doyle,et al.  Modular utility representation for decision-theoretic planning , 1992 .

[3]  R. Reiter,et al.  Embedding decision-analytic control in a learning architecture , 1992 .

[4]  Dimitri P. Bertsekas,et al.  Dynamic Programming: Deterministic and Stochastic Models , 1987 .

[5]  Andrew W. Moore,et al.  The parti-game algorithm for variable resolution reinforcement learning in multidimensional state-spaces , 2004, Machine Learning.

[6]  Sven Koenig,et al.  Optimal Probabilistic and Decision-Theoretic Planning using Markovian , 1992 .

[7]  Peter Haddawy,et al.  Representations for Decision-Theoretic Planning: Utility Functions for Deadline Goals , 1992, KR.

[8]  Leslie Pack Kaelbling,et al.  Planning With Deadlines in Stochastic Domains , 1993, AAAI.

[9]  Stuart J. Russell,et al.  Do the right thing - studies in limited rationality , 1991 .

[10]  R. Howard,et al.  Risk-Sensitive Markov Decision Processes , 1972 .

[11]  John L. Bresina,et al.  Anytime Synthetic Projection: Maximizing the Probability of Goal Satisfaction , 1990, AAAI.

[12]  Sven Koenig,et al.  Utility-Based Planning , 1993 .

[13]  E. Rowland.  Theory of Games and Economic Behavior , 1946, Nature.

[14]  Richard Fikes,et al.  STRIPS: A New Approach to the Application of Theorem Proving to Problem Solving , 1971, IJCAI.

[15]  Richard Goodwin,et al.  Rational Handling of Multiple Goals for Mobile Robots , 1992 .