- aisa agents jo koi b relational action perform kr skta hain, is k liya koi b
percept sequence hoga, jis sequence k accoirding environment ko percieve kiya ja ra
hoga, kio k k definitely jjb hm environemnt sa ksi chz ko sense kr raha hain, to
aik ji dafa completely environment ko sense ni kiya ja skta, to hmara pass
contiously environment sa info recive ho rae hoti h, to jb hm continuosly
environment ko sense kr raha hta hain, to definitely wo sari chzain jin pa hm
sense kr raha hain un sb pa perceptoin generate hoti h, so kbi b ya ni ho skta k at
once sari k sari environment ko sense kr liya jye us k bd aik dafa m tamam
environment pa perception generate ki jye, bl k jsa jsa input aarae hti h wsa hi
perception generate ho ri hti h, to hma ya dkhna pra ga k ksi perticular time pa
agent ka jo percept sequence genrate hwa h , wo sequence kia genrate hwwa h,
Ratioanal agent should select an action for each possible percept sequence, that is
expected to maximize its performnce measure. so environment ko percieve krna k bd
jb b agent apnoi built in knowledge ko use krta hwa koi b action perform kra ga, or
agr wo action relational hoga, to us agent ko hm rational agent kh skta hain.
rationaluty is completely different from omniscience(knowing all thing with
infinite knowledge)
.there is no possiblity k hm aisa agent bnayen jis k pass infinite knowlege ho, kio
k agent m knowledge ko store krana k liya hma storage ki zroort pra gi or aisa
possible hi ni h k hmara pass infinite storage ho, secondaly infinite knowdlege
exist hi ni kerti.
or jtna zyada knowlege agent ko provide kiya ja ra hoga us k liya decision krna
utna hi mushkil hta jye ga. so there is no need of omniscience to develop
rationality.
in rational agents we are interested in autonomous agents. autonomus agents is one
which may have ability to learn and and to adopt the environment. all rational
agents are not autonomous but there is ultimate goal of AI k aisa agents bnaye jye
jo k na srf kam krain bl k environment k according enhance b hon
jb hm rational agents design krna ki bt krta hian to kch important aspects hain
jinha hmain consider krna hta h, or isa hm PEAS k through define kr skta hain.
PEAS is combination of 4 elements:
-Performance measure
-Environemnt
-Actuators
-Sensor
Performance Measure: is k thorogh hm measure krta hain k jo agent hm design kr raha
hain, or us agent ko design krna k liya jo tasks assign kiya hain, kia wo agent un
tasks ko sahi treeka sa pperform kr raha h ya ni kr raha. if we talk abou chess
playing agent to hm bt krain ga k is agent na jo move chli h wo intelligent move
hain ya ni hain.
jo b actions actions pperform kiya ja rae hain legitimate hain k ni
actions legitimate hain lkin kia ya wo actions hain jo k isa task completion
ki trf la k chla jyen, agr action completion ki trf ni la k ja rae to action
rational ni h
Environment: wo environment jis k liya agent ko design kiya ja ra h
-Actuators and sesor ki b zroort hti agent banana k liya
Environment Properties:
there exist mant environment properties:
-fully observable vs partially observable
-deteministic vs stochastic
-episodic vs sequential
-static vs dynamic
-single vs multiple agent
fully observable vs partially observable :
fully obserable hna ka mtlb h k ksi perticular time pa agent environment ko
completely observe kr skta h ya ni kr skta. so agr ksi perticular time pa current
state of environment ko completely obsrve kr skta h to hm kh skta hin k fully
observable h.
lkin agr hmara environment m kch aisa hurdles hn jo k hma completely observe krna
sa rok raha h, ti usa partially observable khta hain.
eg: room cleaning agent is partially obseravable kio k objects hain k jo k sight ko
block kr raha hain jsa class room environemt is also partially obserable h kio k
class room m bht sari hurdles hain
"in fully observable environ,emt agent;s sensor give it access to omplete state of
environment at each point in time."
" in noisy inaccurate sensors or missing parts of state from the sensor data may
lead towards partially observable environment"
deterministic vs stochstis:
determisntic aisa hota h jisma next state of environment is completely dteminie
"the environment is determinsitic if the next state of the environment is
coompletely detemined by the current state and the action executed by the agent"
agr state aik k ilawa or b koi state choose kr skta h, or states dterministic hain
to ya deterministic hna k sath sath stretagic environemnt b hai, jsa k tic tac toe
"if the environment is partially observale the it could appear to be stochastic"
agr environment m other agents k pass aik move k ilawa or koi option na ho to wo
determinstic hta, agr us k pass multiple options hn to wo stretagic b khlata h
stochastic(non determinstic):
if the environment is partially observble then it could be stochas tic environment
but its not mendatory
episodic vs sequential:
