Tweet
Tweet

Slide 1

Slide 1 text

20th Jan. 2023 PhD Defense Planning, Execution, Representation, and Their Integration for Multiple Moving Agents Keisuke Okumura Tokyo Institute of Technology, Japan ౦ژ޻ۀେֶ 5PLZP*OTUJUVUFPG5FDIOPMPHZ http://bit.ly/3krriSO

Slide 2

Slide 2 text

/99 2 YouTube/Mind Blowing Videos logistics YouTube/WIRED manufacturing YouTube/Tokyo 2020 entertainment Swarm { is, will be } necessary in everywhere

Slide 3

Slide 3 text

/99 3 objective-1 Representation objective-2 Planning Common Knowledge? Cooperation? (increased) Uncertainty Execution Navigation for a Team of Agents Who Plans? Huge Search Space

Slide 4

Slide 4 text

/99 4 integration Representation [AAMAS-22+] building effective roadmaps Execution [AAAI-21, ICRA-21, ICAPS-22*, IJCAI-22, AAAI-23+] overcoming uncertainties *Best Student Paper Award Planning ≥1000 agents within 1 sec [IJCAI-19, IROS-21, ICAPS-22*, AIJ-22, AAAI-23+] Research Summary

Slide 5

Slide 5 text

/99 5 Dissertation Overview

Slide 6

Slide 6 text

/99 6 Short-Horizon Planning for MAPF Short-Horizon Planning for Unlabeled-MAPF Short-Horizon Planning Guides Long-Horizon Planning Improving Solution Quality by Iterative Refinement 4. 5. 6. 7. Part I. Planning Building Representation from Learning Building Representation while Planning 11. 12. Part III. Representation Online Planning to Overcome Timing Uncertainties Offline Planning to Overcome Timing Uncertainties Offline Planning to Overcome Crash Faults 8. 9. 10. Part II. Execution Introduction, Preliminaries, and Background 1-3. Conclusion and Discussion 13. Dissertation Outline

Slide 7

Slide 7 text

/99 7 Short-Horizon Planning for MAPF Short-Horizon Planning for Unlabeled-MAPF Short-Horizon Planning Guides Long-Horizon Planning Improving Solution Quality by Iterative Refinement 4. 5. 6. 7. Part I. Planning Dissertation Outline

Slide 8

Slide 8 text

/99 8 Background: Multi-Agent Path Finding (MAPF) given agents (starts) graph goals solution paths without collisions optimization is intractable in various criteria [Yu+ AAAI-13, Ma+ AAAI-16, Banfi+ RA-L-17, Geft+ AAMAS-22]

Slide 9

Slide 9 text

/99 9 Challenge in Planning quality & solvability high low effort small large speed & scalability complete optimal incomplete suboptimal keeping solvability & solution quality with small planning effort holy grail

Slide 10

Slide 10 text

/99 10 Approaches in Planning quality & solvability effort high low small large speed & scalability approach-1 [Chap. 4-6] short-horizon planning guides long-horizon planning approach-2 [Chap. 7] iterative refinement planning time (sec) cost / lower bond 300 agents

Slide 11

Slide 11 text

/99 11 Online Planning to Overcome Timing Uncertainties Offline Planning to Overcome Timing Uncertainties Offline Planning to Overcome Crash Faults 8. 9. 10. Part II. Execution Dissertation Outline

Slide 12

Slide 12 text

/99 12 Challenge in Execution failure demo of synchronous execution planning execution reality gaps overcoming uncertainties at runtime

Slide 13

Slide 13 text

/99 13 Approach in Execution new solution concepts for multi-agent path planning timing-robust conventional 2 1 4 3 0 0 1 2 vulnerable to delays formalize, analyze, and solve problems against: timing uncertainties [Chap. 8-9] / crash faults [Chap. 10]

Slide 14

Slide 14 text

/99 14 Building Representation from Learning Building Representation while Planning 11. 12. Part III. Representation Dissertation Outline

Slide 15

Slide 15 text

/99 15 Background: Planning in Continuous Spaces given agents (starts) goals workspace necessity of constructing roadmap solution paths without collisions

Slide 16

Slide 16 text

/99 16 Challenge in Representation dense representation sparse representation produced by PRM [Kavraki+ 96] complete optimal incomplete suboptimal building small but promising representation for planning quality & solvability effort high low small large speed & scalability

Slide 17

Slide 17 text

/99 17 Approaches in Representation quality & solvability effort high low small large speed & scalability approach-1 [Chap. 11] learning from planning demonstrations approach-2 [Chap. 12] building representation while planning

Slide 18

Slide 18 text

/99 18 Short-Horizon Planning for MAPF Short-Horizon Planning Guides Long-Horizon Planning 4. 6. Part I. Planning Building Representation from Learning Building Representation while Planning 11. 12. Part III. Representation Rest of Talk: Quick Multi-Agent Path Planning discrete continuous

Slide 19

Slide 19 text

/99 19 Short-Horizon Planning for MAPF Short-Horizon Planning Guides Long-Horizon Planning 4. 6. Part I. Planning Building Representation from Learning Building Representation while Planning 11. 12. Part III. Representation Quick Multi-Agent Path Planning discrete continuous

Slide 20

Slide 20 text

/99 20 MAPF: Multi-Agent Path Finding given agents (starts) graph goals solution paths without collisions optimization is intractable in various criteria [Yu+ AAAI-13, Ma+ AAAI-16, Banfi+ RA-L-17, Geft+ AAMAS-22]

Slide 21

Slide 21 text

/99 21 quality & solvability effort high low small large speed & scalability keeping solvability & solution quality with small planning effort complete optimal incomplete suboptimal keeping solvability & solution quality with small planning effort Challenge in Planning

Slide 22

Slide 22 text

/99 22 solved instances (%) runtime (sec) Results on MAPF Benchmark [Stern+ SOCS-19] 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC 33 grid maps 0.0% A* [Hart+ 68] example 194x194 |V|=13,214

Slide 23

Slide 23 text

/99 23 solved instances (%) runtime (sec) 0.0% A* [Hart+ 68] 0.4% ODrM* [W agner+ AIJ-15] complete & optimal Results on MAPF Benchmark [Stern+ SOCS-19] 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC

Slide 24

Slide 24 text

/99 24 solved instances (%) runtime (sec) 0.0% A* [Hart+ 68] 0.4% ODrM* [W agner+ AIJ-15] 8.3% CBS [Sharon+ AIJ-15; Li+ AIJ-21] 10.7% BCP [Lam+ COR-22] Results on MAPF Benchmark 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC solution complete* & optimal *SC: solution complete: unable to distinguish unsolvable instances comp & optimal

Slide 25

Slide 25 text

/99 25 solved instances (%) runtime (sec) 0.0% A* [Hart+ 68] 0.4% ODrM* [W agner+ AIJ-15] 8.3% CBS [Sharon+ AIJ-15; Li+ AIJ-21] 10.7% BCP [Lam+ COR-22] 30.9% ODrM*-5 [W agner+ AIJ-15] Results on MAPF Benchmark [Stern+ SOCS-19] 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC comp & bounded sub-opt SC* & optimal *SC: solution complete: unable to distinguish unsolvable instances comp & optimal

Slide 26

Slide 26 text

/99 26 solved instances (%) runtime (sec) 0.0% A* [Hart+ 68] 0.4% ODrM* [W agner+ AIJ-15] 8.3% CBS [Sharon+ AIJ-15; Li+ AIJ-21] 10.7% BCP [Lam+ COR-22] 30.9% ODrM*-5 [W agner+ AIJ-15] 50.5% EECBS-5 [Li+ AAAI-21] Results on MAPF Benchmark [Stern+ SOCS-19] 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC comp & bounded sub-opt SC & bounded sub-opt SC* & optimal *SC: solution complete: unable to distinguish unsolvable instances comp & optimal

Slide 27

Slide 27 text

/99 27 solved instances (%) runtime (sec) 0.0% A* [Hart+ 68] 0.4% ODrM* [W agner+ AIJ-15] 8.3% CBS [Sharon+ AIJ-15; Li+ AIJ-21] 10.7% BCP [Lam+ COR-22] 30.9% ODrM*-5 [W agner+ AIJ-15] 50.5% EECBS-5 [Li+ AAAI-21] 61.4% PP [Silver AIIDE-05] 80.9% LNS2 [Li+ AAAI-22] comp & bounded sub-opt SC & bounded sub-opt incomp & sub-opt SC* & optimal comp & optimal Results on MAPF Benchmark [Stern+ SOCS-19] 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC [Stern+ SOCS-19] *SC: solution complete: unable to distinguish unsolvable instances

Slide 28

Slide 28 text

/99 28 long-horizon (deliberative, offline) planning stage acting sage short-horizon (reactive, online) Unblock Me, Google Play Planning Horizon

Slide 29

Slide 29 text

/99 29 Strategy in Planning quality & solvability effort high low small large speed & scalability complete optimal incomplete suboptimal long-horizon (deliberative, offline) short-horizon (reactive, online)

Slide 30

Slide 30 text

/99 30 Strategy in Planning quality & solvability effort high low small large speed & scalability integration short-horizon planning guides long-horizon planning long short

Slide 31

Slide 31 text

/99 31 Proposed MAPF Algorithms quality & solvability effort high low small large speed & scalability PIBT LaCAM LaCAM PIBT

Slide 32

Slide 32 text

/99 32 Proposed MAPF Algorithms quality & solvability effort high low small large speed & scalability PIBT LaCAM

Slide 33

Slide 33 text

/99 33 planning online planning applicable to lifelong scenarios ≥500 agents within 50ms ensuring that all agents eventually reach their destinations https://kei18.github.io/pibt2 IJCAI-19 => AIJ-22 Priority Inheritance with Backtracking for Iterative Multi-agent Path Finding KO, Manao Machida, Xavier Defago & Yasumasa Tamura scalable sub-optimal algorithm (PIBT) to solve MAPF iteratively

Slide 34

Slide 34 text

/99 34 locations at t=1 t=2 t=3 repeat one-timestep prioritized planning high low mid How PIBT works – 1/6 … 1 2 3 4 5 6 7 8 9 decision order time-window

Slide 35

Slide 35 text

/99 35 How PIBT works – 2/6 simple prioritized planning is incomplete high low mid stuck

Slide 36

Slide 36 text

/99 36 How PIBT works – 3/6 high low mid as high priority inheritance [Sha+ IEEE Trans Comput-90]

Slide 37

Slide 37 text

/99 37 high low mid How PIBT works – 4/6 1 3 2 decision order … …

Slide 38

Slide 38 text

/99 38 How PIBT works – 5/6 high as high as high as high as high stuck but still not feasible

Slide 39

Slide 39 text

/99 39 How PIBT works – 6/6 invalid valid re-plan re-plan valid You can move invalid You must re-plan, I will stay introduce backtracking repeat this one-timestep planning until termination

Slide 40

Slide 40 text

/99 40 Theoretical Result With dynamic priorities,* in biconnected graphs, all agents reach their destinations within finite timestep convenient in lifelong scenarios important note: PIBT is incomplete for MAPF unsolvable MAPF, but the theorem holds *s.t. unreached agents eventually get the highest

Slide 41

Slide 41 text

/99 41 prioritized planning w/distance-based heuristics [Silver AIIDE-05] A* with operator decomposition greedy version [Standley AAAI-10] EECBS CBS-based, bounded sub-optimal [Li+ AAAI-21] LNS2 large neighborhood search for MAPF [Li+ AAAI-22] PIBT Performance on one-shot MAPF 25 instances 30sec timeout on desktop PC sufficiently long timestep limit 194x194 four-connected grid sum-of-costs (normalized; min:1) runtime (sec) success rate agents not so bad not so bad blazing fast! worst: 550ms

Slide 42

Slide 42 text

/99 42 PIBT is great but… agents room-64-64-8 64x64 |V|=3,232 ost003d 194x194 |V|=13,214 agents random-32-32-20 32x32 |V|=819 agents important note: PIBT is incomplete for MAPF We need one more jump! PIBT 0% success rate in 30sec

Slide 43

Slide 43 text

/99 43 Proposed MAPF Algorithms quality & solvability effort high low small large speed & scalability PIBT LaCAM

Slide 44

Slide 44 text

/99 44 planning LaCAM: Search-Based Algorithm for Quick Multi-Agent Pathfinding KO agents room-64-64-8 ost003d agents random-32-32-20 agents success rate in 30sec 100% LaCAM worst: 699ms worst: 394ms worst: 11sec quick & complete algorithm for MAPF (LaCAM; lazy constraints addition search) https://kei18.github.io/lacam AAAI-23

Slide 45

Slide 45 text

/99 45 … … … … … search node (configuration) Vanilla A* for MAPF greedy search: 44 nodes in general: (5^N)xT nodes N: agents, T: depth intractable even with perfect heuristics goal configuration

Slide 46

Slide 46 text

/99 46 PIBT for MAPF PIBT PIBT PIBT greedy search: 44 nodes only 4 configurations repeat one-timestep planning until termination use PIBT to guide exhaustive search initial configuration goal configuration

Slide 47

Slide 47 text

/99 47 … … … … … Concept of LaCAM PIBT PIBT PIBT use other MAPF algorihtms to generate a promising configuration configurations are generated in a lazy manner by two-level search scheme exhaustive search but node generation are dramatically reduced => quick & complete MAPF

Slide 48

Slide 48 text

/99 48 Lazy Constraints Addition constraint tree (maintained implicitly) invoked multiple times during the search

Slide 49

Slide 49 text

/99 49 1st invoke configuration generation with Lazy Constraints Addition no constraint

Slide 50

Slide 50 text

/99 50 Lazy Constraints Addition 2nd invoke configuration generation with

Slide 51

Slide 51 text

/99 51 Lazy Constraints Addition e.g., breadth-first search 24th invoke Each configuration eventually generates all connected configurations. Theorem: LaCAM is complete discard node from Open configuration generation with

Slide 52

Slide 52 text

/99 52 sum-of-costs (normalized; min:1) runtime (sec) success rate agents prioritized planning w/distance-based heuristics [Silver AIIDE-05] A* with operator decomposition greedy version [Standley AAAI-10] EECBS CBS-based, bounded sub-optimal [Li+ AAAI-21] LNS2 large neighborhood search for MAPF [Li+ AAAI-22] PIBT [Okumura+ AIJ-22] Performance on one-shot MAPF 25 instances, on laptop 30sec timeout sufficiently long timestep limit 194x194 four-connected grid not so bad (diff from PIBT: tie-breaking) LaCAM blazing fast! worst: 699ms perfect!

Slide 53

Slide 53 text

/99 53 success rate in 1000sec runtime (sec) agents (x1000) 25 instances, timeout of1000 sec, on desctop PC warehouse-20-40-10-2-2, 340x164; |V|=38,756 Performance with x1000 agents perfect! blazing fast! worst: 26sec for 10,000 agents LaCAM wrong thought: “centralized algorithms are not scalable” => NO, the game is changing PIBT demo of 10,000 agents

Slide 54

Slide 54 text

/99 54 planning Improving LaCAM for Scalable Eventually Optimal Multi-Agent Pathfinding KO (under review) LaCAM*: complete algorithm eventually converging to optima improved configuration generator: PIBT with reversed vertex scoring configuration & cost 1 1 2 3 4 5 6 0 before 1 1 2 3 3 2 3 0 after rewriting new edge updated by Dijkstra

Slide 55

Slide 55 text

/99 55 solved instances (%) runtime (sec) 0.0% A* [Hart+ 68] 0.4% ODrM* [W agner+ AIJ-15] 8.3% CBS [Sharon+ AIJ-15; Li+ AIJ-21] 10.7% BCP [Lam+ COR-22] 30.9% ODrM*-5 [W agner+ AIJ-15] 50.5% EECBS-5 [Li+ AAAI-21] 61.4% PP [Silver AIIDE-05] 80.9% LNS2 [Li+ AAAI-22] Results on MAPF Benchmark 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC comp & bounded sub-opt SC & bounded sub-opt incomp & sub-opt SC* & optimal comp & optimal [Stern+ SOCS-19] *SC: solution complete: unable to distinguish unsolvable instances

Slide 56

Slide 56 text

/99 56 solved instances (%) runtime (sec) 67.4% PIBT [Okumura+ AIJ-22] blazing fast! But room for improvements 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC comp & bounded sub-opt SC & bounded sub-opt incomp & sub-opt SC* & optimal comp & optimal Results on MAPF Benchmark [Stern+ SOCS-19] *SC: solution complete: unable to distinguish unsolvable instances

Slide 57

Slide 57 text

/99 57 solved instances (%) runtime (sec) complete & eventually optimal! comp & optimal 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC 99.0% LaCAM* blazing fast! comp & bounded sub-opt SC & bounded sub-opt incomp & sub-opt SC* & optimal comp & optimal Results on MAPF Benchmark [Stern+ SOCS-19] *SC: solution complete: unable to distinguish unsolvable instances

Slide 58

Slide 58 text

/99 58 solved instances (%) runtime (sec) comp & bounded sub-opt SC & bounded sub-opt incomp & sub-opt SC* & optimal comp & optimal Results on MAPF Benchmark [Stern+ SOCS-19] remaining: maze-128-128-1** **I do not care; it will be too optimized for the benchmark 13900 instances - 33 grid maps - random scenario - every 50 agents, up to max. (1000) tested on desktop PC 99.0% LaCAM* complete & eventually optimal! 32/33 maps: all solved in 10sec *SC: solution complete: unable to distinguish unsolvable instances

Slide 59

Slide 59 text

/99 59 Summary in Planning quality & solvability effort high low small large speed & scalability short-horizon planning guides long-horizon planning PIBT LaCAM(*) successfully breaking the trade-off

Slide 60

Slide 60 text

/99 60 Short-Horizon Planning for MAPF Short-Horizon Planning Guides Long-Horizon Planning 4. 6. Part I. Planning Building Representation from Learning Building Representation while Planning 11. 12. Part III. Representation Quick Multi-Agent Path Planning discrete continuous

Slide 61

Slide 61 text

/99 61 MAPF Definition Again given agents (starts) graph goals solution paths without collisions reality is continuous

Slide 62

Slide 62 text

/99 62 Cool coordination but not efficient! why not move diagonally? robots follow grid [Okumura+ ICAPS-22]

Slide 63

Slide 63 text

/99 63 Unsolvable Instance? obstacle

Slide 64

Slide 64 text

/99 64 Multi-Agent Path Planning in Continuous Spaces given agents (starts) goals solution paths without collisions finding solutions itself is tremendously challenging [Spirakis+ 84, Hopcroft+ IJRR, Hearn+ TCS-05] workspace

Slide 65

Slide 65 text

/99 65 artificial potential field sampling-based rule-based goal start Strategies to Solve Single-Agent Path Planning in Continuous Spaces constructing roadmap

Slide 66

Slide 66 text

/99 66 SBMP: Sampling-Based Motion Planning state of robot: (x, y) (x, y) should be in this region configuration space random sampling & construct roadmap pathfinding on roadmap same scheme even in high-dimension

Slide 67

Slide 67 text

/99 67 Naïve Strategy to Solve Multi-Agent Path Planning in Continuous Spaces Construct agent-wise roadmaps by SBMP (sampling-based motion planning) methods Solve MAPF on those roadmaps 1. 2. two-phase planning

Slide 68

Slide 68 text

/99 68 Pitfall – There is a trade-off dense representation sparse representation produced by PRM [Kavraki+ 96] complete optimal incomplete suboptimal ideal: small but promising representation for planning quality & solvability effort high low small large speed & scalability

Slide 69

Slide 69 text

/99 69 Countermeasure biased sampling sampling from important region of each agent how to identify? agent-specific features + interactions between agents 🤔 design manually?

Slide 70

Slide 70 text

/99 70 Countermeasure biased sampling sampling from important region of each agent how to identify? agent-specific features + interactions between agents This is machine learning problem! supervised learning: planning demonstration as training data

Slide 71

Slide 71 text

/99 71 quality & solvability effort high low small large speed & scalability learning from planning demonstrations CTRMs: Learning to Construct Cooperative Timed Roadmaps for Multi-agent Path Planning in Continuous Spaces KO,* Ryo Yonetani, Mai Nishimura & Asako Kanezaki representation https://omron-sinicx.github.io/ctrm AAMAS-22 *work done as an intern at OMRON SINIC X MAPF algorithm

Slide 72

Slide 72 text

/99 72 𝐹!"#$ model training instances & solutions predict next locations MAPF algorithm new instance 𝐹!"#$ random walk sampling module next locations for all agents starts path generation compositing solution … t=0 t=1 t=2 CTRMs Workflow Online Inference Offline Training CVAE: Conditional Variational Autoencoder [Sohn+ NeurIPS-15] +importance sampling [Salzmann+ ECCV-20] +multi-agent attention [Hoshen NeurIPS-17] 𝐹"#$% :

Slide 73

Slide 73 text

/99 𝐹"#$% 73 next position instance & solution occupancy cost-to-go env. info Offline Training & Model Arch. [Sohn+ NeurIPS-15] CVAE features ?

Slide 74

Slide 74 text

/99 𝐹"#$% 74 + + goal-driven features relative positions, size, speeds, etc Offline Training & Model Arch.

Slide 75

Slide 75 text

/99 𝐹"#$% 75 + + comm. features attention Offline Training & Model Arch. [Hoshen NeurIPS-17]

Slide 76

Slide 76 text

/99 𝐹"#$% 76 go right [0,0,1] indicator feature Offline Training & Model Arch.

Slide 77

Slide 77 text

/99 𝐹"#$% 77 + + + + go right [0,0,1] next position goal-driven features comm. features indicator feature instance & solution occupancy cost-to-go env. info relative positions, size, speeds, etc attention Offline Training & Model Arch. [Sohn+ NeurIPS-15] CVAE [Hoshen NeurIPS-17]

Slide 78

Slide 78 text

/99 78 observations for agent-i next predicted location for agent-i trained model likely to be used by planners Online Inference

Slide 79

Slide 79 text

/99 79 Online Inference timestep t timestep t+1 next predicted locations for all agents observations for all agents

Slide 80

Slide 80 text

/99 80 Online Inference 0 1 2 T T-1 … initial locations timed path for agent-i each path is agent-specific and cooperative hyperparameter

Slide 81

Slide 81 text

/99 81 … … … … compositing 0 1 2 T T-1 timed roadmap for agent-i each roadmap is agent-specific and cooperative hyperparameter: #(path generation) Online Inference

Slide 82

Slide 82 text

/99 82 SPARS [Dobson & Bekris, IJRR-14] (random) simplified PRM [Karaman & Frazzoli, IJRR-11] square as agent-specific roadmaps grid as used in MAPF studies CTRMs 20-30 homogeneous agents corresponding to 32x32 grids Roadmap Visualization CTRMs produce small but effective roadmaps specific to each agent

Slide 83

Slide 83 text

/99 83 Quantitative Results S R P R R R 50 RP 3 5 T 20-30 homogeneous agents corresponding to 32x32 grids 100 instances solved by prioritized planning [Silver, AIIDE-05, Van Den Berg & Overmars IROS-05, etc] CTRMs reduce planning effort while keeping solution qualities params of CTRMs: #(path generations) denser denser

Slide 84

Slide 84 text

/99 84 50 RP 3 5 T P R R P S S 20-30 homogeneous agents corresponding to 32x32 grids 100 instances solved by prioritized planning [Silver, AIIDE-05, Van Den Berg & Overmars IROS-05, etc] CTRMs achieve efficient path-planning from the end-to-end perspective Roadmap construction can be much faster. Check our latest implementation: https://github.com/omron-sinicx/jaxmapp Quantitative Results quick! denser

Slide 85

Slide 85 text

/99 85 Strategy in Representation quality & solvability effort high low small large speed & scalability learning from planning demonstrations building representation while planning

Slide 86

Slide 86 text

/99 86 Naïve Strategy to Solve Multi-Agent Path Planning in Continuous Spaces Are roadmaps truly used in planning process? Construct agent-wise roadmaps by SBMP (sampling-based motion planning) methods Solve MAPF on those roadmaps 1. 2. two-phase planning decoupled

Slide 87

Slide 87 text

/99 87 Advanced Strategy to Solve Multi-Agent Path Planning in Continuous Spaces develop agent-wise roadmaps by SBMP according to MAPF search progress coupled

Slide 88

Slide 88 text

/99 88 Quick Multi-Robot Motion Planning by Combining Sampling & Search KO & Xavier Defago (under review) planning representation https://kei18.github.io/sssp algorithm (SSSP) to solve multi-robot motion planning quickly by simultaneously performing roadmap construction & collision-free pathfinding

Slide 89

Slide 89 text

/99 89 MRMP: Multi-Robot Motion Planning MAPF is a special case of MRMP solution: collision-free trajectories each agent has its own configuration space blackbox utility functions connect free space true false steer collide true false sample dist 0.18

Slide 90

Slide 90 text

/99 90 Proposed Algorithm: SSSP 0 1 2 0 1 2 3 0 1 2 3 4 5 0 1 2 2 3 5 0 1 2 3 4 0 1 4 00 00 10 20 40 50 00 50 51 53 54 00 10 20 40 50 00 vertex expansion & goals new vertices closed next action exhaustive search while constructing roadmaps by random walks +many tricks inspired by SBMP & MAPF studies [MAPF; Standley AAAI-10] [SBMP; Hsu+ ICRA-97] search-node expansion

Slide 91

Slide 91 text

/99 91 Theoretical Result SSSP eventually finds a solution for solvable instances c.f., probabilistic completeness in SBMP

Slide 92

Slide 92 text

/99 92 R P 3 0 6 33 6663 Point2d DOF: 2N Point3d DOF: 3N R P 33 6 6663 Line2d DOF: 3N R P 3 0 33 6 6663 Capsule3d DOF: 6N R P 3 0 33 6 6663 Arm22 DOF: 2N Arm33 DOF: 6N R P 3 0 33 6 6663 Dubins2d DOF: 3N R P 6 33 6663 Snake2d DOF: 6N Performance of SSSP very promising quick!

Slide 93

Slide 93 text

/99 93 Summary in Representation quality & solvability effort high low small large speed & scalability learning from planning demonstrations building representation while planning successfully breaking the trade-off again

Slide 94

Slide 94 text

/99 94 Closing

Slide 95

Slide 95 text

/99 TODAY 95 Planning Representation Execution integration [AAMAS-22+] building effective roadmaps [AAAI-21, ICRA-21, ICAPS-22, IJCAI-22, AAAI-23+] overcoming uncertainties ≥1000 agents within 1 sec [IJCAI-19, IROS-21, ICAPS-22, AIJ-22, AAAI-23+] Research Summary

Slide 96

Slide 96 text

/99 96 Impact of Research Results breaking common belief of trade-off (quality vs speed) connecting discrete & continuous spaces => promoting various types of automation develop new horizons of quick multi-agent path planning contribution:

Slide 97

Slide 97 text

/99 97 YouTube/Mind Blowing Videos warehouse YouTube/StarCraft video game [Flatland Challenge, AIcrowd] railway/airport operations [Morris+ AAAIW-16] [Le Goc+ UIST-16] robotic interface [Song+ ICCBB-01] drug discovery [Zhang+ 18] manufacturing [van den Berg+ ISRR-11] crowd simulation [Zhang+ SIGGRAPH-20] puzzle solving pipe routing [Belov+ SOCS-20] Impact of Research Results develop new horizons of quick multi-agent path planning contribution:

Slide 98

Slide 98 text

/99 98 Future Direction integration LaCAM(*) PIBT planning in discretized spaces SSSP CTRMs planning in continuous spaces establish practical methodologies to MRMP iterative refinement robust execution scalable & quick probabilistically complete asymptotically optimal, anytime with kinodynamic constraints

Slide 99

Slide 99 text

/99 More Info? => Check My Website! https://kei18.github.io/ Thank You for Listening!

Slide 100

Slide 100 text

/99 100 Publications 1. “Improving LaCAM for Scalable Eventually Optimal Multi-Agent Pathfinding.” KO (under review) 2. “Quick Multi-Robot Motion Planning by Combining Sampling and Search.” KO & Xavier Défago. (under review) 3. “LaCAM: Search-Based Algorithm for Quick Multi-Agent Pathfinding.” KO. AAAI. 2023. 4. “Fault-Tolerant Offline Multi-Agent Path Planning.” KO & Sebastien Tixeuil. AAAI. 2023. 5. “Priority Inheritance with Backtracking for Iterative Multi-agent Path Finding.” KO, Manao Machida, Xavier Défago & Yasumasa Tamura. Artificial Intelligence (AIJ). 2022 (previously presented at IJCAI-19) 6. “Offline Time-Independent Multi-Agent Path Planning.” KO, François Bonnet, Yasumasa Tamura & Xavier Défago. IJCAI. 2022. (extended version is under review at T-RO) 7. “Solving Simultaneous Target Assignment and Path Planning Efficiently with Time-Independent Execution.” KO & Xavier Défago. ICAPS. 2022. (best student paper award, extended version is under review at AIJ) 8. “CTRMs: Learning to Construct Cooperative Timed Roadmaps for Multi-agent Path Planning in Continuous Spaces.” KO, Ryo Yonetani, Mai Nishimura & Asako Kanezaki. AAMAS. 2022. 9. “Iterative Refinement for Real-Time Multi-Robot Path Planning.” KO, Yasumasa Tamura & Xavier Défago. IROS. 2021. Others: 10. “Active Modular Environment for Robot Navigation.” Shota Kameyama, KO, Yasumasa Tamura & Xavier Défago. ICRA. 2021. 11. “Time-Independent Planning for Multiple Moving Agents.” KO, Yasumasa Tamura & Xavier Défago. AAAI. 2021. 12. “Roadside-assisted Cooperative Planning using Future Path Sharing for Autonomous Driving.” Mai Hirata, Manabu Tsukada, KO, Yasumasa Tamura, Hideya Ochiai & Xavier Défago. VTC. 2021. 13. “winPIBT: Extended Prioritized Algorithm for Iterative Multi-agent Path Finding.” KO, Yasumasa Tamura & Xavier Défago. WoMAPF. 2021. 14. “Amoeba Exploration: Coordinated Exploration with Distributed Robots.” KO, Yasumasa Tamura & Xavier Défago. iCAST. 2018. under review peer-reviewed