Markov Decision Processes
PPTs by Dan Klein and Pieter Abbeel for CS188 Intro to AI at UC Berkeley.
Non-Deterministic Search
Example: Grid World
▪ A maze-like problem
▪ The agent lives in a grid
▪ Walls block the agent’s path
▪ Noisy movement: actions do not always go as planned
▪ 80% of the time, the action North takes the agent North
(if there is no wall there)
▪ 10% of the time, North takes the agent West; 10% East
▪ If there is a wall in the direction the agent would have
been taken, the agent stays put
▪ The agent receives rewards each time step
▪ Small “living” reward each step (can be negative)
▪ Big rewards come at the end (good or bad)
▪ Goal: maximize sum of rewards
Grid World Actions
Deterministic Grid World Stochastic Grid World
Markov Decision Processes
▪ An MDP is defined by:
▪ A set of states s ∈ S
▪ A set of actions a ∈ A
▪ A transition function T(s, a, s’)
▪ Probability that a from s leads to s’, i.e., P(s’| s, a)
▪ Also called the model or the dynamics
▪ A reward function R(s, a, s’)
▪ Sometimes just R(s) or R(s’)
▪ A start state
▪ Maybe a terminal state
▪ MDPs are non-deterministic search problems
▪ One way to solve them is with expectimax search
▪ We’ll have a new tool soon
[Demo – gridworld manual intro (L8D1)]
Video of Demo Gridworld Manual Intro
What is Markov about MDPs?
▪ “Markov” generally means that given the present state, the
future and the past are independent
▪ For Markov decision processes, “Markov” means action
outcomes depend only on the current state
Andrey Markov
(1856-1922)
▪ This is just like search, where the successor function could only
depend on the current state (not the history)
Policies
▪ In deterministic single-agent search problems,
we wanted an optimal plan, or sequence of
actions, from start to a goal
▪ For MDPs, we want an optimal policy π*: S → A
▪ A policy π gives an action for each state
▪ An optimal policy is one that maximizes
expected utility if followed
▪ An explicit policy defines a reflex agent
Optimal policy when R(s, a, s’) = -0.03
▪ Expectimax didn’t compute entire policies for all non-terminals s
▪ It computed the action for a single state only
Optimal Policies
R(s) = -0.01 R(s) = -0.03
R(s) = -0.4 R(s) = -2.0
Example: Racing
Example: Racing
▪ A robot car wants to travel far, quickly
▪ Three states: Cool, Warm, Overheated
▪ Two actions: Slow, Fast
▪ Going faster gets double reward 0.5 +1
1.0
Fast
Slow -10
+1
0.5
Warm
Slow
Fast 0.5 +2
Cool 0.5
+1 Overheated
1.0
+2
Racing Search Tree
MDP Search Trees
▪ Each MDP state projects an expectimax-like search tree
s s is a state
(s, a) is a
s, a
q-state
(s,a,s’) called a transition
s,a,s’ T(s,a,s’) = P(s’|s,a)
R(s,a,s’)
’s
Utilities of Sequences
Utilities of Sequences
▪ What preferences should an agent have over reward sequences?
▪ More or less? [1, 2, 2] or [2, 3, 4]
▪ Now or later? [0, 0, 1] or [1, 0, 0]
Discounting
▪ It’s reasonable to maximize the sum of rewards
▪ It’s also reasonable to prefer rewards now to rewards later
▪ One solution: values of rewards decay exponentially
Worth Now Worth Next Step Worth In Two Steps
Discounting
▪ How to discount?
▪ Each time we descend a level, we
multiply in the discount once
▪ Why discount?
▪ Sooner rewards probably do have
higher utility than later rewards
▪ Also helps our algorithms converge
▪ Example: discount of 0.5
▪ U([1,2,3]) = 1*1 + 0.5*2 + 0.25*3
▪ U([1,2,3]) < U([3,2,1])
Stationary Preferences
▪ Theorem: if we assume stationary preferences:
▪ Then: there are only two ways to define utilities
▪ Additive utility:
▪ Discounted utility:
Quiz: Discounting
▪ Given:
▪ Actions: East, West, and Exit (only available in exit states a, e)
▪ Transitions: deterministic
▪ Quiz 1: For γ = 1, what is the optimal policy?
▪ Quiz 2: For γ = 0.1, what is the optimal policy?
▪ Quiz 3: For which γ are West and East equally good when in state d?
Infinite Utilities?!
▪ Problem: What if the game lasts forever? Do we get infinite rewards?
▪ Solutions:
▪ Finite horizon: (similar to depth-limited search)
▪ Terminate episodes after a fixed T steps (e.g. life)
▪ Gives nonstationary policies (π depends on time left)
▪ Discounting: use 0 < γ < 1
▪ Smaller γ means smaller “horizon” – shorter term focus
▪ Absorbing state: guarantee that for every policy, a terminal state will eventually
be reached (like “overheated” for racing)
Recap: Defining MDPs
▪ Markov decision processes: s
▪ Set of states S
▪ Start state s0 a
▪ Set of actions A s, a
▪ Transitions P(s’|s,a) (or T(s,a,s’))
▪ Rewards R(s,a,s’) (and discount γ) s,a,s’
’s
▪ MDP quantities so far:
▪ Policy = Choice of action for each state
▪ Utility = sum of (discounted) rewards
Solving MDPs
Optimal Quantities
▪ The value (utility) of a state s:
V*(s) = expected utility starting in s and s s is a
acting optimally state
a
(s, a) is a
▪ The value (utility) of a q-state (s,a): s, a q-state
Q*(s,a) = expected utility starting out
having taken action a from state s and s,a,s’ (s,a,s’) is a
(thereafter) acting optimally ’s transition
▪ The optimal policy:
π*(s) = optimal action from state s
[Demo – gridworld values (L8D4)]
Snapshot of Demo – Gridworld V Values
Noise = 0
Discount = 1
Living reward = 0
Snapshot of Demo – Gridworld Q Values
Noise = 0
Discount = 1
Living reward = 0
Snapshot of Demo – Gridworld V Values
Noise = 0.2
Discount = 1
Living reward = 0
Snapshot of Demo – Gridworld Q Values
Noise = 0.2
Discount = 1
Living reward = 0
Snapshot of Demo – Gridworld V Values
Noise = 0.2
Discount = 0.9
Living reward = 0
Snapshot of Demo – Gridworld Q Values
Noise = 0.2
Discount = 0.9
Living reward = 0
Snapshot of Demo – Gridworld V Values
Noise = 0.2
Discount = 0.9
Living reward = -0.1
Snapshot of Demo – Gridworld Q Values
Noise = 0.2
Discount = 0.9
Living reward = -0.1
Values of States
▪ Fundamental operation: compute the (expectimax) value of a state
▪ Expected utility under optimal action
s
▪ Average sum of (discounted) rewards
▪ This is just what expectimax computed! a
s, a
▪ Recursive definition of value:
s,a,s’
’s
Racing Search Tree
Racing Search Tree
Racing Search Tree
▪ We’re doing way too much
work with expectimax!
▪ Problem: States are repeated
▪ Idea: Only compute needed
quantities once
▪ Problem: Tree goes on forever
▪ Idea: Do a depth-limited
computation, but with increasing
depths until change is small
▪ Note: deep parts of the tree
eventually don’t matter if γ < 1
Time-Limited Values
▪ Key idea: time-limited values
▪ Define Vk(s) to be the optimal value of s if the game ends
in k more time steps
▪ Equivalently, it is what a depth-k expectimax would give from s
[Demo – time-limited values (L8D6)]
Noise = 0.2
Discount = 0.9
Living reward = 0
Consider Grid world: Sates are presented
as (Row, Colum)
Possible actions at each state: Move one
step in either North, South, East, West
k=0
Noise = 0.2
Discount = 0.9
Living reward = 0
For K=0, V0(S) is optimal value of a state
S, if game ends in zero time steps
i.e. No time to get any expected reward
thus zero utility (value)
Therefore, at k=0, every state has 0 value
k=0
Noise = 0.2
Discount = 0.9
Living reward = 0
k=1
Noise = 0.2
Discount = 0.9
Living reward = 0
k=2
Noise = 0.2
Discount = 0.9
Living reward = 0
k=3
Noise = 0.2
Discount = 0.9
Living reward = 0
k=4
Noise = 0.2
Discount = 0.9
Living reward = 0
k=5
Noise = 0.2
Discount = 0.9
Living reward = 0
k=6
Noise = 0.2
Discount = 0.9
Living reward = 0
k=7
Noise = 0.2
Discount = 0.9
Living reward = 0
k=8
Noise = 0.2
Discount = 0.9
Living reward = 0
k=9
Noise = 0.2
Discount = 0.9
Living reward = 0
k=10
Noise = 0.2
Discount = 0.9
Living reward = 0
k=11
Noise = 0.2
Discount = 0.9
Living reward = 0
k=12
Noise = 0.2
Discount = 0.9
Living reward = 0
k=100
Noise = 0.2
Discount = 0.9
Living reward = 0
Computing Time-Limited Values
Value Iteration
Value Iteration
▪ Start with V0(s) = 0: No time steps left means an Vk+1(s)
expected reward sum of zero
a
▪ Given vector of Vk(s) values, s, a
s,a,s’
▪ Do one ply of expectimax from every state: Vk(s’)
▪ Repeat until convergence
Value Iteration
▪ Complexity of each iteration: O(S2A) Vk+1(s)
▪ Theorem: will converge to unique optimal values
a
s, a
▪ Basic idea: approximations get refined
towards optimal values s,a,s’
▪ Policy may converge long before values do Vk(s’)
Example: Value Iteration
Initially : V0(s) = 0: No time steps left means an expected reward sum of zero
0 0 0
States
Cool Warm Overheated
Example: Value Iteration
Now, Calculate V1(s)
Using value update equation:
Assume no discount! i.e. γ = 1
Available actions in various states:
Cool: Slow, Fast; Warm: Slow, Fast; Overheated: None
0 0 0
States
Cool Warm Overheated
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
2 1 0
0 0 0
States
Cool Warm Overheated
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
Example: Value Iteration
As we assume no discount i.e. γ = 1
Value updates will continue for ever (will not
3.5 2.5 0 converse)
2 1 0
0 0 0
States
Cool Warm Overheated
Example: Value Iteration
However, if . γ < 1
It will converse.
3.5 2.5 0
Stopping Criteria: When states values become
stable i.e. almost no change in value for next
time step.
2 1 0
0 0 0
States
Cool Warm Overheated
Noise = 0.2
Discount = 0.9
Living reward = 0
Consider Grid world: Sates are presented
as (Row, Colum)
Possible actions at each state: Move one
step in either North, South, East, West
k=0
Noise = 0.2
Discount = 0.9
Living reward = 0
For k = 0, V0(S) is optimal value of a state
S, if game ends in zero time steps
i.e. No time to get any expected reward
thus zero utility (value)
Therefore, at k = 0, every state has 0 value
k=1
Value update of states after 1 time-step:
For k = 1, V1(S)
In one time-step (k = 1): There is NO reward to get
for any states, except states (1, 3) and (2, 3),
In state (1, 3) robot gets -1 reward and exit the game,
In state (2, 3) robot gets +1 reward and exit the game.
k=1
(2, 0)
(2, 0)
k=1 (2, 0) (2, 1)
(1 0)
(2, 0)
k=1 (2, 0) (2, 1)
(1 0)
(2, 0)
k=1 (2, 0) (2, 1)
(1 0)
(2, 0)
k=1 (2, 0) (2, 1)
(1 0)
(2, 0)
k=1 (2, 0) (2, 1)
(1 0)
(2, 0)
k=1
(2, 2)
(2, 2)
k=1 (2, 1) (2, 2) (2, 3)
(1, 2)
(2, 0)
k=1 (2, 1) (2, 2) (2, 3)
(1, 2)
(2, 0)
k=1 (2, 1) (2, 2) (2, 3)
(1, 2)
(2, 0)
k=1 (2, 1) (2, 2) (2, 3)
(1, 2)
(2, 0)
k=1 (2, 1) (2, 2) (2, 3)
(1, 2)
(2, 0)
k=1
Value update of states after 1 time-step:
For k = 1, V1(S)
For both Sates (1, 3) and (2, 3): Only action available
is (2, 2)
Available action: Exit
Thus, No need to calculate Q values as no uncertainty.
V1 ((1, 3)) = -1 and V1 ((2, 3)) = +1
k=2
(2, 0)
k=2
(2, 0)
k=2
(2, 0)
k=2
(2, 0)
k=2
(2, 0)
k=2
(2, 0)
k=2
(2, 2)
k=2 (2, 1)
(2, 0)
k=2 (2, 1)
(2, 0)
k=2
(2, 0)
k=2
(2, 0)
k=2
(2, 0)
k=2
(1, 2)
k=2
(2, 0)
k=2
(2, 0)
k=2
(2, 0)
k=2
(2, 0)
k=2
(2, 0)
k=2
Value update of states after 1 time-step:
For k = 2, V2(S)
For both Sates (1, 3) and (2, 3): Only action available
is (2, 2)
Available action: Exit
Thus, No need to calculate Q values as no uncertainty.
V2 ((1, 3)) = -1 and V2 ((2, 3)) = +1
k=2
Noise = 0.2
Discount = 0.9
Living reward = 0
k=3
Noise = 0.2
Discount = 0.9
Living reward = 0
k=4
Noise = 0.2
Discount = 0.9
Living reward = 0
k=5
Noise = 0.2
Discount = 0.9
Living reward = 0
k=6
Noise = 0.2
Discount = 0.9
Living reward = 0
k=7
Noise = 0.2
Discount = 0.9
Living reward = 0
k=8
Noise = 0.2
Discount = 0.9
Living reward = 0
k=9
Noise = 0.2
Discount = 0.9
Living reward = 0
k=10
Noise = 0.2
Discount = 0.9
Living reward = 0
k=11
Noise = 0.2
Discount = 0.9
Living reward = 0
k=12
Noise = 0.2
Discount = 0.9
Living reward = 0
k=100
Noise = 0.2
Discount = 0.9
Living reward = 0
Convergence*
▪ How do we know the Vk vectors are going to converge?
▪ Case 1: If the tree has maximum depth M, then VM holds
the actual untruncated values
▪ Case 2: If the discount is less than 1
▪ Sketch: For any state Vk and Vk+1 can be viewed as depth
k+1 expectimax results in nearly identical search trees
▪ The difference is that on the bottom layer, Vk+1 has actual
rewards while Vk has zeros
▪ That last layer is at best all RMAX
▪ It is at worst RMIN
▪ But everything is discounted by γk that far out
▪ So Vk and Vk+1 are at most γk max|R| different
▪ So as k increases, the values converge
