A Novel Congruent Organizational Design Methodology Using Group Technology and a Nested Genetic Algorithm

Transcription

1 A Novel Congruent Organizational Design Methodology Using Group Technology and a Nested Genetic Algorithm Feili Yu Georgiy M. Levchuk Candra Meirina Sui Ruan Krishna Pattipati 9-th International CCRTS, June, 2004 Track 1: Modeling and Simulation

2 Overview Motivation Problem Formulation Solution Approach Group Technology and Nested GA Performance Measures Numerical Simulation Conclusion

3 Motivation - 1 Design Process for Command Organizations Mission Multi-Dimensional Task Decomposition Organizational Constraints Resource Capabilities Organizational Design Process 1) Task - DM Allocation 2) Resource - DM Allocation 3) Resource Task Allocation Quantitative Mission Structure Task Requirements FEED-BACK FEED-BACK Organizational Structure - Functional Allocation (allocation pattern of tasks, resources, and DMs) -Coordination Strategy (coordination pattern) -Task Processing Strategy (Task sequencing) Organizational Performance Criteria

4 Motivation - 2 Current Three-phase Design Methodology Operational Task Requirements Operational Resource Capabilities Task Precedence Graph Algorithms: DLS, PWE, Roll-out phase 1 Task Task Schedule(s); Schedule(s); Resource Resource Packages Packages Task Task - - Resource Resource Allocation Allocation Decision Decision Rules Rules DM-Task DM-Task Assignment; Assignment; DM s DM s Operational Operational Workload Workload Algorithms: Hierarchical Clustering phase 2 DM DM - - Resource Resource Allocation Allocation DM s Expertise Constraints feedback DM s DM s Coordination Coordination Load; Load; DM s DM s Responsibilities Responsibilities Algorithms: Min-max, Max-span, Min-coord DM DM Hierarchy Hierarchy phase 3 Communication Constraints

5 Motivation - 3 Drawbacks of 3-Phase Design Methodology The organizational design problem has been decomposed into several sub-problems to overcome computational complexity Phase I does not account for the workload of inter-dm coordination, which may cause high degree of sub-optimality in phases II and III The 3-phase design process does not take into account the task execution accuracy; it assumes that all the task requirements can be fully satisfied, which is not true in practice

6 Motivation - 4 Mission responsibility assignment task allocation: Functional organization: Assets/resources of the same type Mission responsibility by functional area Divisional organization: Assets/resources of different types Mission responsibility by geographical area What organization lies between functional and divisional? Hybrid responsibility rules?

7 Motivation - 5 geography Example of hybrid assignment by decision trees: Region A Region B Divisional Functional yes no yes no A R 1 >R 2 Tasks in A Tasks in B Tasks with higher R 1 Tasks with higher R 2 Resources: R 1( = strike), R2( = AAW ) DM1 DM2 DM1 DM2 resources geography Hybrid T T T T T T T R R 2 4 DM DM 1 3 DM T 1 Region A T 2 Region B T 7 T 6 T 5 DM1 T 1,T 2 yes yes A T 3 T 4 DM2 DM1 DM2 DM2 DM1 no no yes R 1 >1 R 2 <2,T 2 3,, T, T 3,,, no T T,T 6 7 7

8 Problem Formulation - 1 The objective is to minimize the aggregated workload of each DM, which takes into account both the intra-dm and the inter-dm coordination workloads. In order to balance the workloads among DMs, we seek to minimize the root mean-square value of the aggregated workload, which is given by: Minimize: WRMS = 1 M M m= 1 W 2 ( m) where W ( m) = αw ( m) + (1 α ) W ( m) Intra Inter (m) W Intra (m) W Inter α is the intra-dm workload of DM m is the inter-dm workload is the weight assigned to the intra-dm workload Subject to: 1. Each Platform can only be assigned to one DM 2. Each DM should be assigned at least one Task

9 Problem Formulation - 2 The objective function (1) can be separated into 2 sub-problems: 1. Minimize the intra-dm workload W = Intra ( m) ρ1 T T TAS i T T [ AIntra ( m, i)] i m t( m, i) = i m t( m, i) where t( m, i) is the overall platform transfer time when processing task T i in DM m A Intra ( m, i) is processing accuracy of task T i in DM m (2) ρ 1 is the intra-dm task accuracy significance index 2. Minimize the inter-dm workload W Inter = M m= 1 W Inter ( m) = T Tˆ i [ Aˆ tˆ( i) Inter ( i)] ρ 2 (3) where t(i) A Inter ρ 2 (i) is the inter-dm platform transfer time when processing coordination task T i is task accuracy of coordination task T i is the inter-dm task accuracy significance index

10 Solution: Group Technology - 1 What is Group Technology (GT)? Group technology (GT) recognizes and exploits similarities in three distinct ways: by performing similar operations together by standardizing similar tasks by efficiently storing and retrieving information about recurring problems GT can be carried out by dividing a C 2 system into several manageable subsystems or cells, responsible for managing tasks, assets (platforms), and information flow

11 Solution: Group Technology - 2 The advantages of introducing GT into C 2 systems are: Improved speed of command Reduced task latencies (execution delays) Reduced resource requirements Reduced mission inefficiencies Reduced synchronization delays Reduced response time Improved flexibility Deconfliction - identifying responsibility areas GT algorithms: Matrix-based Clustering Hierarchical Clustering Graph-Theoretic Clustering AI based Clustering Evolutionary Clustering Decision-tree Clustering

12 Example (1) Example Task resources requirement data: Platform capability data: Task ID Task Name AAW ASUW ASW GASLT FIRE ARM MINE DES Locations Pro. Times 1 CVBG ARG Resupply Port North Resupplu Port South Encounters North&South HILL NORTH BEACH SOUTH BEACH Defend N. Beach Defend S. Beach S/P Road A/P Road SAM SeaPort SAM AirPort SEAPORT AIRPORT GTL Blow Bridge Platform ID Platform Name AAW ASUW ASW GASLT FIRE ARM MINE DES Velocity 1DDG FFG CG ENG INFA SD AHI CAS CAS CAS VF VF VF SMC TARP SAT SOF INF(AAAV-1) INF(AAAV-2) INF(MV22-1) requirements/capabilities are modeled: AAW (Anti-Air Warfare), ASUW (Anti- Surface Warfare), ASW (Anti-Submarine Warfare), GASLT (Ground Assault), FIRE (Artillery), ARM (Armor), MINE (Mine Clearing), DES (Designation)

13 Example (2) Platforms/Assets Tasks Cluster OUTSIDE: tasks & platforms Cluster INSIDE: tasks to platforms P5 P6 DM-platform-task allocation DM1 DM2 DM3 DM3 DM4 Nested Grouping Process

14 Example (3) Tasks Tasks Platforms Clustering (NGA) Platforms Tasks-platform relationship before clustering Tasks-platform grouping after clustering DM1 Ground Assault DM2 DM3 DM4 Attack Defend Marine Operations DM Functionality DM-DM Coordination Structure

15 Nested GA Procedure The Nested GA is comprised of two loops: Outer-loop and Inner-loop There are two stages for the Inner-loop: Inner-loop1 and Inner-loop2 Initial taskplatform grouping Unfinished tasks and related Platforms Cost of grouping and task-platform assignment Outer-Loop GA Inner-Loop 1 Inner-Loop 2 Initial Population Group 1 Group 2 Group M Run GA to update the task-platform assignment Matrix Crossover and mutate parents to produce new generations 1. Number of DMs 2. Task/Platforms Grouping 3. Platform-Task Assignment 4. Aggregated Workload Inner-Loop GA New Generations

16 Performance Measures A. Average Platform Transfer Time Total intra-dm and inter-dm transfer time of platforms divided by number of platforms B. Clustering Efficiency Ratio of task-platform assignment in groups to the total task-platform assignment C. Average Task Accuracy Sum of each task accuracy over number of tasks. Average Task Accuracy measures how good the overall tasks have been processed D. Average Platform Utilization Sum of utilization of each platform over number of platforms. The utilization of each platform is the percentage of resource capability of platform being used for task execution

17 Conclusion Introduced Group Technology (GT) concept into organizational design Proposed a two-layer algorithm framework for solving organizational design problem Applied Nested GA (NGA) as a solution approach Defined performance measures Numerical simulation shows that this solution approach is capable of designing a congruent organization in terms of resource and task allocation structure Next step: Implement decision-tree clustering