Title: Propagating Uncertainty In POMDP Value Iteration with Gaussian Process
1Propagating Uncertainty In POMDP Value Iteration
with Gaussian Process
Written by Eric Tuttle and Zoubin
Ghahramani Presenter by Hui Li May 20, 2005
2- Outline
- Framework of POMDP
- Framework of Gaussian Process
- Gaussian Process Value Iteration
- Results
- Conclusions
3Framework of POMDP
The POMDP is defined by the tuple lt S, A, T, R,
?,O gt
- S is a finite set of states of the world.
- A is a finite set of actions.
- T S?A ? ?(S) is the state-transition function,
the probability of an action changing the the
world state from one to another,T(s, a, s). - R S?A ? ? is the reward for the agent in a
given world - state after performing an action, R(s, a).
- ? is a finite set of observations.
- O S?A ? ?(?) is the observation function, the
probability of making a certain observation
after performing a particular action, landing in
state s, O(s, a, o).
4POMDP agent can be decomposed into two parts a
state estimator (SE) and a policy (?).
o
b
a
5The goal of a POMDP agent is to select actions
which maximize its expected total sum of future
rewards .
- Two functions are most often used in
reinforcement learning algorithms - value function (function of state)
Optimal value function
- Q function (function of state-action)
Optimal value function
6The key assumption of POMDP is that the state is
unknown, partially observable. We rely on the
concept of a belief state, denoted b, to
represent a probability distribution over states.
The belief is a sufficient statistic for a given
history
7After taking an action a and seeing an
observation o, the agent updates its belief state
using Bayes rule
8Bellmans equations for POMDP are as follows
Bellmans equations for value function
Bellmans equations for Q function
9Framework of Gaussian Process regression
A Gaussian process regressor defines a
distribution over possible functions that could
fit the data. In particular, the distribution of
a function y(x) is a Gaussian process if the
probability density p(y(x1), y(x2), , y(xN)) for
any finite set of points x1,, xN is a
multivariate Gaussian.
10Assume we have a Gaussian process with mean 0 and
covariance function K(xi, xj).
Suppose we have observed a set of training points
and target function values D (xn,tn), n 1,,
N.
? N(0, ?), a Gaussian noise. Then C ? K.
With a new data x, we have
11One general choice of covariance function is
With W a diagonal matrix.
? expected amplitude of the function ? a bias
term that accommodates non-zero-mean functions.
Using maximum likelihood or MAP methods, the
parameters of the covariance function can be
tuned.
12Gaussian Process Value Iteration
Q function
Model each of the action value functions Q(? ,a)
as a Gaussian process. According to the
definition of the Gaussian process, Qt-1(? ,a)
is a multivariate normal distribution with mean
?a,bo and covariance ?a,bo.
The major problem in computing the distribution
Qt(b ,a) is the max operator.
13Two approximate ways for dealing with the max
operator
1. Approximate the max operator as simply passing
through the random variable with the highest
mean.
Where
142. Take into account the effects of the max
operator, but ignore correlations among the
function values.
If q1 and q2 are independent with distributions
Then first two moments of variable q max( q1,
q2) are given
Where
? is the cdf and ? is the pdf for a zero mean,
unit variance normal.
And q can be approximate using a Gaussian
distribution.
15Based on that, we can use a Gaussian distribution
to approximate
Both methods produce a Gaussian approximation for
the max of a set of normally distributed vectors.
And since Qta is related to Qt-1 by a linear
transformation , we have
16Results
17Conclusions
- In this paper, authors presented an algorithms
Gaussian processes for approximate value
iteration in POMDPs. - The results using GP are comparable to that of
the classical methods.