P LANNING WITH I NCOMPLETE U SER P REFERENCES AND D OMAIN M ODELS Tuan Anh Nguyen Graduate Committee Members: Subbarao Kambhampati (Chair) Chitta Baral Minh B. Do Joohyung Lee David E. Smith
M OTIVATION Automated Planning In practice… Research: Action models are not Actions available upfront Preconditions Cost of modeling Effects Error-prone Deterministic Users usually don’t Non-deterministic exactly know what Stochastic Initial situation they want Goal conditions Always want to see more than one plan What a user wants about plans Find a (best) plan! 2 Planning with incomplete user preferences and domain models
Preferences in Planning – Traditional View Classical Model: “Closed world” assumption about user preferences. All preferences assumed to be fully specified/available Full Knowledge of Preferences Two possibilities If no preferences specified — then user is assumed to be indifferent. Any single feasible plan considered acceptable. If preferences/objectives are specified, find a plan that is optimal w.r.t. specified objectives. Either way, solution is a single plan 3 3
Preferences in Planning — Real World Full Knowledge Real World: Preferences not fully known of Preferences is lacking Unknown preferences For all we know, user may care about every thing --- the flight carrier, the arrival and departure times, the type of flight, the airport, time of travel and cost of travel… Partially known We know that users cares only about travel time and cost. But we don’t know how she combines them… 4 4
Domain Models in Planning – Traditional View Classical Model: “Closed world” assumption about action descriptions. Full Knowledge Fully specified preconditions and effects of domain models Known exact probabilities of outcomes pick-up :parameters (?b – ball ?r – room) :precondition (and (at ?b ?r) (at-robot ?r) (free-gripper)) :effect (and (carry ?b) (not (at ?b ?r)) (not (free-gripper))) 5 5
Domain Models in Planning – (More) Practical View Completely modeling the domain dynamics Time consuming Error-prone Sometimes impossible What does it mean by planning with incompletely specified domain models? Plan could fail! Prefer plans that are more likely to succeed… How to define such a solution concept? 6 6
Problems and Challenges Incompleteness representation Solution concepts Planning techniques 7 7
D ISSERTATION OVERVIEW “Model - lite” Planning Preference Domain incompleteness incompleteness Representation Representation Solution concept Solution concept Solving techniques Solving techniques 8
D ISSERTATION OVERVIEW “Model - lite” Planning Preference Domain incompleteness incompleteness Representation: two levels Representation of incompleteness Actions with possible User preferences exist, but preconditions / effects totally unknown Optionally with weights Partially specified for being the real ones Complete set of plan Solution concept: “robust” attributes plans Parameterized value function, unknown Solving techniques: trade-off values synthesizing robust plans Solution concept: plan sets Solving techniques: 9 synthesizing high quality plan sets
D ISSERTATION O VERVIEW “Model - lite” Planning Preference incompleteness Distance measures w.r.t. Representation: two levels base-level features of plans of incompleteness (actions, states, causal links) User preferences exist, but CSP-based and local-search totally unknown based planners Partially specified Full set of plan attributes IPF/ICP measure Parameterized value Sampling, ICP and Hybrid function, unknown approaches trade-off values Solution concept: plan sets with quality Solving techniques: 10 synthesizing quality plan sets
D ISSERTATION OVERVIEW “Model - lite” Planning Preference Domain incompleteness incompleteness Publication Publication Representation: two levels Representation of incompleteness Domain independent approaches Assessing and Generating Actions with possible for finding diverse plans . IJCAI Robust Plans with Partial User preferences exist, but preconditions / effects (2007) Domain Models . ICAPS-WS totally unknown Optionally with weights (2010) Planning with partial preference Partially specified for being the real ones models. IJCAI (2009) Synthesizing Robust Plans under Full set of plan Incomplete Domain Models. Solution concept: “robust” Generating diverse plans to attributes AAAI-WS(2011), NIPS (2013) plans handle unknown and partially Parameterized value known user preferences. AIJ 190 A Heuristic Approach to function, unknown Solving techniques: (2012) Planning with Incomplete trade-off values synthesizing robust plans STRIPS Action Models. ICAPS (with Biplav Srivastava, Subbarao Solution concept: plan sets (2014) Kambhampati, Minh Do, Alfonso (with Subbarao Kambhampati, Solving techniques: Gerevini and Ivan Serina) 11 Minh Do) synthesizing high quality plan sets
P LANNING WITH I NCOMPLETE D OMAIN M ODELS 12
R EVIEW : STRIPS Predicate set R : clear(x – object), on- table(x – object), on(x – object, y – object), holding(x – object), hand-empty Operators O : Name (signature): pick-up(x – object) Preconditions: hand-empty, clear(x) Effects: ~hand-empty, holding(x), ~clear(x) A single complete model! 13
P LANNING P ROBLEM WITH STRIPS Set of typed objects {𝑝 1 , … , 𝑝 𝑙 } Together with predicate set 𝑄 , we have a set of grounded propositions 𝐺 Together with operators 𝑃 , we have a set of grounded actions 𝐵 Initial state: 𝐽 ∈ 𝐺 Goals: 𝐻 ⊆ 𝐺 14
P LANNING P ROBLEM WITH STRIPS (2) Find : a plan 𝜌 achieves 𝐻 starting from 𝐽 : 𝛿 𝜌, 𝐽 ⊇ 𝐻. Transition function: 𝑏 , 𝑡 = 𝑡 ∪ 𝐵𝑒𝑒 𝑏 ∖ 𝐸𝑓𝑚(𝑏) for applying 𝑏 ∈ 𝐵 𝛿 in 𝑡 ⊆ 𝐺 s.t. 𝑄𝑠𝑓 𝑏 ⊆ 𝑡 . Applying 𝜌 = 〈𝑏 1 , … , 𝑏 𝑜 〉 at state 𝑡 : 𝛿 𝜌, 𝑡 = 𝛿( 𝑏 𝑜 , 𝛿( 𝑏 2 , … , 𝑏 𝑜−1 , 𝑡)) 15
I NCOMPLETE D OMAIN M ODELS Predicate set 𝑺 : clear(x – object), on-table(x – object), on(x – object, y – object), holding(x – object), hand- empty, light(x – object), dirty(x – object) Operators 𝑷 Name (signature): pick-up(x – object) Incompleteness Preconditions: hand-empty, clear(x) in deterministic Possible preconditions: light(x) domains Effects: ~hand-empty, holding(x), ~clear(x) Possible effects: dirty(x) Stochastic domains = 〈𝑺, 𝑷〉 Incomplete domain 𝑬 At “schema” level with typed variables (no objects) With K “annotations”, we have 2 𝐿 possible complete models, one of which is the true model . 16
P LANNING P ROBLEM WITH I NCOMPLETE DOMAIN Set of typed objects {𝑝 1 , … , 𝑝 𝑙 } Together with predicate set 𝑄 , we have a set of grounded propositions 𝐺 Together with operators 𝑃 , we have a set of grounded actions 𝐵 Initial state: 𝐽 ∈ 𝐺 Goals: 𝐻 ⊆ 𝐺 Find : a plan 𝜌 “achieves” 𝐻 starting from 𝐽 An ill-defined solution concept! 17 Need a definition for “goal achievement”
T RANSITION F UNCTION , applying 𝝆 in s results in a set of possible Under 𝑬 states: 𝜹 𝑬 𝒋 (𝝆, 𝒕) 𝜹 𝝆, 𝒕 = ≫ 𝑬 𝒋 ∈≪𝑬 The probability of ending up in 𝒕 ′ ∈ 𝜹(𝝆, 𝒕) is equal to 𝑸𝒔(𝑬 𝒋 ) ≫, 𝒕 ′ =𝜹 𝑬𝒋 (𝝆,𝒕) 𝑬 𝒋 ∈≪𝑬 where 𝑸𝒔 (𝑬 𝒋 ) is the probability of 𝑬 𝒋 being the true 18 model.
T RANSITION F UNCTION 𝜹 𝑬 ( 𝒃 , 𝒕) : STRIPS Execution (SE): 𝐸 (〈𝑏〉, 𝑡) = 𝑡 ∖ 𝐸𝑓𝑚 𝐸 𝑏 ∪ 𝐵𝑒𝑒 𝐸 𝑏 , 𝑗𝑔 𝑄𝑠𝑓 𝐸 𝑏 ⊆ 𝑡 𝑝𝑢ℎ𝑓𝑠𝑥𝑗𝑡𝑓 𝛿 𝑇𝐹 𝑡 ⊥ = {⊥} , 𝑄𝑠𝑓 𝐸 𝑏 ⊈ 𝑡 ⊥ , 𝐻 ⊈ 𝑡 ⊥ ⊥ ∉ 𝐺 Generous Execution (GE): 𝐸 (〈𝑏〉, 𝑡) = 𝑡 ∖ 𝐸𝑓𝑚 𝐸 𝑏 ∪ 𝐵𝑒𝑒 𝐸 𝑏 , 𝑗𝑔 𝑄𝑠𝑓 𝐸 𝑏 ⊆ 𝑡 𝛿 𝐻𝐹 𝑡, 𝑝𝑢ℎ𝑓𝑠𝑥𝑗𝑡𝑓 20
Initial state 𝐽 = {𝑞 2 } Proposition set 𝐺 = {𝑞 1 , 𝑞 2 , 𝑞 3 } 21 Goal 𝐻 = {𝑞 3 }
A M EASURE FOR P LAN R OBUSTNESS Naturally, we prefer plan that succeeds in as many complete models as possible 𝑆 𝜌 = |Π| 2 𝐿 𝑺 𝑻𝑭 𝝆 ≤ 𝑺 𝑯𝑭 (𝝆) R GE 𝜌 = 6/8 R GE 𝜌 = 4/8 22
A BIT MORE GENERAL … Predicate set 𝑺 : clear(x – object), on-table(x – object), on(x – object, y – object), holding(x – object), hand-empty, light(x – object), dirty(x – object) Operators 𝑷 Name (signature): pick-up(x – object) Preconditions: hand-empty, clear(x) Possible preconditions: light(x) with a weight of 0.8 Effects: ~hand-empty, holding(x), ~clear(x) Possible effects: dirty(x) with an unspecified weight Treat weights as probabilities with random variables Robustness measure: 𝑺 𝝆 ≝ 𝐐𝐬 (𝑬 𝒋 ) 23 〉〉:𝜹 𝑬𝒋 𝝆,𝑱 ⊨𝑯 𝑬 𝒋 ∈ 〈〈𝑬
Recommend
More recommend