9 Assignment Problems Learning Objectives, The learning objectives of this chapter are to; © to introduce the assignment model and the requirements to solve the assignment problems to demonstrate the Hungarian algorithm to solve the assignment prob- lems © to introduce variants of assignment problems and demonstrate the solution procedure for those variants pee koncs The learners will be able to: © prepare assignment models and solve them, @ solve and interpret the solutions of different variants of the assignment problems © model and solve transshipment problems Assignment Problem considers the assignment of a group of tasks to be carried out by exactly the same number of workers, groups. machines, etc. Each task has a one-to-one assignment with workers or facilities. It is also applicable to fit people or tasks to different time slots to achieve benefits in the form of reduced costs. time or effort. Even though the general objective in the assignment problem is to reduce the sociated total cost components, it is equally applicable to maximization problems through slight modification of the problem. ‘This type of problems involve matching of items in a set of resources to another item in a set of tasks and vice versa such as: task-worker, job-machine, time-person, lime-group, position-person, job-time, ete. The requirement in this type of problems270 Operations Research: Concepts, Problems & Solu, ions is that the matching has to be established one-to-one, Therefore, to assign my humbep of tasks. there should be exactly m number of resources. In the cases of unequal rn unbalanced problem, The solution procedure for suey numbers, it is called a1 9.2.2. problems are discussed in Section 9.1. Problem Structure and Assumptions The assignment problems in general have the following structure. There are 2 pool of resources or people and another pool of tasks or jobs. Each resource. task pair has an associated cost. Only one of these pairs will be selected for cach resource end for each task leading to the final solution. The variable in the ifies the binary condition whether a pair selected or not, esource j. the variable x,, becomes true or 1. All other feta atly. for any task k # i, x4; = 0. vs made based on the cost components in every cell of 2 the associated costs in the respective cells. Rows and columns represent the resturces/people and the jobs/tasks, respectively. be shown interchanged, The entries in the ssignments. If the number +s usually denoted problem si sk i is assigned sociated in various pos by mii | ws the number of tas viz. m =n, necessery modificeions are made + equal before solving. Figure 9.) shows easigreent problercs. Sect cctvAssignment Problerna on 3 The number of resource/facility is equal to the number of jobaltzake end usually denoted by n. 4 A resource/facility, say , assigned for a job, say J”. will cause 2 cont of. 5 The decision varinbles are x,, thar will take assigned to j” facility, x will 6 Total number of decision varizbles =n 7 n 7 Every (i, j) pair will have 4 cont associated with it, ot of assignment in the position, (i, jy becomes: Cy. values 10, J}. Sf the & job is ,otherwine 0. hay Which is either O oF Cy 9 The objective is to minimize the total cost: EE cis, 10 Two sets of constraints have to be satisfied while solving the problems: (a) the total nurnber of allocations for a resource/facility (rerw wire ellccationsy has to be one, and (b) total number of allocations for 2 job (column wise allocations) has to be one. That is, A iss Of the assignment problem is as follows number of row constraint equation all of them defined over 2 set of n* number of Minimize, Z=) Yeux, es) fal pA subject to. Diyeh 120 (92) ra Mayak W=12 (93) 4 =00rk: Hi. jy (Ay 91.2 Cost Matrix Cost matrix of assignment problem is a squase matrix of size n 2s shown in Figure 9.2. The available n resources are assigned to the n tasks thes provides the Minimum pensible total cow. The resources are shown 2s B, to B, end the tasks are shown as Ay to Ag.Operations Research: Concepts, Problems & Solution, Is Ay Ap Aj + An 272 Rp] en | ez | aj foes | Sin Ro | ai | €22 | C27 [or Can Ri} ci | Cig | Cij fore | Cin Rn | em | ena | nj fo | enn Figure 9.2: Assignment problem cost matrix 9.1.3 Types of Assignment Models Assignment problems are of different types. Though the general objective is minimizing cost. there are problems of maximizing objective. The matrix in such problems represent the benefits of different assignments rather than the costs. Such problems have to be converted to minimization type by the method similar to that adopted for maximization transportation problem conversion. Unbalanced problems are another type of problems. The number of resources and the jobs are different in such problem. From a pool of resources, the optimum ones have to be selected for the tasks. In the other setting, the best suitable jobs/tasks are allotted to the resources that minimizes the cost. 9.2 Solution to Assignment Problems The solution to the problem can be prepared by enumeration if the matrix size is small. For an mx n matrix, the number of sequences possible to check becomes n!. These assignment sequences with the total cost have to be compared to get the minimum cost assignment. For a matrix of size 4, the number of such possible assignments become 24. Up to a size of 3, this enumeration process is manageable by hand. Check the following example. + Example 9.1 Assume a cost matrix of an assignment problem as follows. 3 machines M1, M2, and M3 have to be assigned to the available 3 jobs, J1, J2, and J3. Each job when performed in these machines take different time to complete. Determine the minimum cost assignment to the problem. Table shows time in hours. Jobs no2 B 2M 2 6 3 . 3 M2 3° 5 4 2 M3 5 4 6 al‘Assignment Problems os «Solution 9.1 In this small problem there are six alternate assignments possible. ‘They can be shown as follows. M/c Job Assignment Mi non Rn Rp B M2 23 3B WN NW I M3 3°22 U0 B 2 Time(hrs) 13° 10 15 15 10 13 Out of these enumeration, two assignments are found to have the minimum possible duration of 10 hours. The assignments are: M1-J1; M2-J3; M3-J2, and MI-J3; M2-J1; M3-J2. These two assignments will give the same duration as observed. . 9.2.1 Hungarian Assignment Algorithm Enumeration method fails as the matrix size increases. For a 3x5 matrix itself the enumeration required becomes 120 and with 5 entries in each assignment makes it more difficult to handle, Hungarian Assignment Algorithm provides a systematic procedure to solve assignment problems of higher size. The steps in Hungarian algorithm are as follows. . Write down the cost of assignments as a square matrix of size n. If not square. add dummy columns (rows) of zero cost to make it square. . Subtract the minimum element in each row from all elements in the respective row. 3. From the resultant matrix, subtract minimum element in every column from all elements of each column if all columns do not have an element as zero. 4. Draw minimum number (N) of lines to cover alll the zeros. There can be N=norN + Solution 96 The problem gives the runs as benefits from every assignment of batsman to the different batting positions, The maximum runs by any batsman in any position is 49. Therefore, all entries in the matrix will be subtracted from this maximum run 49: Fig, 9.14(a). After the row and column operations the matrix changes as: Fig. 9.14(b). The zeros in the resultant matrix ca be of five lines. Therefore, a fe: rows have one completely covered by a minimum asible solution exists, The first row and the fourth ero cach. Therefore, they can be assigned first and then the other ignments also can be made without any arbitrary selection, Fig. 9.14(c) shows the assignments made in the matrix which is unique, ‘The optimal assignment of batsmen to the positions are: A - V, B= 1V, C= 1th D-1.E- I. Therefore, the batting order is: D-E-C-B-A, The expected runs from these batsmen can be found out by selecting the runs from these assigned cells of the original matrix. ‘The expected runs in this batting order = 49 +394 444 464482226 *~ ‘pgsignmont Problems 283 gli | 1/0 16] 11/3} 1] 0 16} 11] 3 | 1 {Lo 1017 6|0|5}0}7 6|4\5 |[o] 7 19] 15 0} 1] 0 }14}10 W| 1 |[o]j14}10 11|16 0/4 6] 8/13 o|] 4| 6| 8 \13 7/9 3/0/0}6/8 3 [fol] #| 6] 8 (a) Minimization matrix (b) Row & column operation (c) Optimal solution Figure 9.14: Solution to Example 9.6 9.4 Crew Scheduling Problem Crew assignment arises in the transport sector. Some examples are: Airlines Crew scheduling, Freight Crew scheduling, bus crew scheduling, etc. In these examples, the crew needs to be provided with proper layover time. For this, the crew should be stationed at some location such that the schedules can be operated without any problem of insufficient layover. There arise the pairing of schedules to the crew in onward and return journeys. In such problems, the objective will be to minimize the total layover time for the crew. Different pairing of schedules and stationing of crew at different cities will provide layover time for the crew. The minimization of the layover times are achieved by the Hungarian assignment algorithm. Example 9.7 An airline that operates seven days a week has the following time table. The crew must have a minimum layover of 5 hours between flights. Obtain the pairing of flights that minimizes crew layover time away from home. For any pairing, the crew will be based at the city that provides smaller layover. Flight | Chennai-Bangalore | Flight | Bangalore-Chennai | No Depart Arrive No. Depart Arrive CBI | 7:00am 8:00am | BCI 8:00 am 9:15 am CB2 | 8:00am 9:00am | BC2 $:30am 9:45 am CB3 | 1:30pm 2:30pm | BC3 | 12:00noon 1:15 pm CB4 [6:30pm 7:30pm | BC4 | 5:30pm __ 6:45 pm For each pair also mention in which city the crew must be stationed. . = Solution 9.7 The crew that depart from stationed city can be assumed to have a pairing flight from the other city to come back. Similarly, if the crew is stationed in the other city, the crew will reach this city and return to the station city in another nearest flight. The crew should however be allowed with a minimum layover time in the transit city. The problem should be solved considering two possibilities of station city for the same pair of flights, Example; Flight No. 201 and 101 can be paired by assuming the station city either as Delhi or as Jaipur. Both will give different layover times. These layover times are found out consider the two cities separately for all the flightaN ones Concepts, Problems S § Solutions over for sae pang ad the corresponding station ation, sill be pained iin layover consider ng minimim erence between the arrival time at the tansit city till matrix and for both the cities. ring benween 201 and 101, with the crew ity, the flight reaches at 8:00 am. only on the next day at 8:00 am. 24 hours. utes: “(83 5 he). The layover matrices can be obtained as: I | Bcl | BC? | BC3 | BCS | [CBI 17.75 | 12.25 | (cB? | 75 [1 | (ces | 2825 | 24.25 [18.75 | [eps | 925 | 8 75 | Siationed at Chennai ch pairing. select the minimum time from the nwo tables. This minimum jer the minimum required time for the crew layover in each city, Here is ven as 5 brs. Therefore, the selection of the minimum layover will be above hours minimum. This selection can happen from both the city as the station city. The entries will be multiplied by 4 to remove the decimal portion of the time Values. Using * for Chennai. the minimum layover time matrix becomes. Bcl | BC? | BCS | BC | [Bci | BC2 | BC3 | BC+ 21.73" | 2 [1775 | 95 _| cBi | s7= | ss* | 7i= | 38 | [ep por | aor | 75* | 34 | | cB3 | 70 | 72 | 86 | 75* | [eps | 37 | 35 | 21 | 88 ‘Affer row and column ‘operations. the matrix changes tor = BCl BC2_BC3_ BCA] BC1| BC2| BC3| BC4 CB 49" | 47" 33" | | [epil ase | a5 | 33* [0 cB2_ 57" 58" 41" | 0 | | CB2| 5z* | 53* | ae | 0 cas 0 2 | 6 | | ces) 0 | 0 | 6 | ces ie" 14" | or | 67 |) cBa) er | 2" | o | 67 18.75"Assignmont Problems The zeros in the matrix can be covered by three lines. Therefore, a feasible solution does not exist, From the uncovered cells, the minimum 33 is subtracted and added to the crossing cell elements, BCI] BC2| BC3| BCa BCI cB1| 49° | 45" | 33" | 6 | | CB1| 16" | 12° | o | 0 cB2| 57* | 53* | 41" |) 0 cB2| 24" | 20° | 8° o | b €B3} -0 -+ -0 -+ 16-+ 5*-| | CB3| 0 0 | 16 | 38* Fepat as aa*-- -o-F 67-| | cB] 6" | 12° | o* | 100 in three lines can cover all the zeros. Therefore. the minimum of uncovered jements subtracted from all uncovered cells and added to the crossing cells. BCl] BC2| BC3] BC4) BC1| BC2| BC3| BC4) ci] 16* | 12* | o | 0 cpi| 4° | o | ot | 0 | cB2| 24* | 20% | 8 [ 0 cB2| 12* | 8* | 8* | 0 | | 2 t €B3 | o--a6-- 38+] | cB3] 0 | 0 | 28 | 50°) cB4| 16* | 12* | o* | 160 CB4| 4* o* o* 100 | Now the minimum number of lines required to cover all the zeros is four. Two feasible solutions exist as one of the zeros to be chosen arbitrarily. The solutions are: BCI] BC2] BC3| BC4| BCi| BC2| BC3| BC4| cB1| 4* | [of | w | a cpl] 4a | a | jor] a | cp2| 12° | 8= | 8* | [ol | | cB2| 12* | a« | 8° | Jo] | cps] [o] | a | 28 | so*| | cp3| [oj] | w | 28 | 50] cpa] 4° | we | fof | 100 || cBal 4* | fos | we | 100) The results have to be interpreted as follows. Solution 1: CB1-BC2 to be paired and crew to be stationed at Chennai (* for Chennai): 21.25 hrs. CB2-BC4 to be paired and crew to be stationed at Bangalore: 8.5 hrs. CB3-BCI to be paired and crew to be stationed at Bangalore: 17.5 hrs. CB4-BC3 to be paired and crew to be stationed at Chennai: 5.25 hrs. Total layover time = 52 hrs. 30 minutes Solution 2: CB1-BC3 to be paired and crew to be stationed at Chennai (* for Chennai): 17.75 hrs. CB2-BC4 to be paired and crew to be stationed at Bangalore: 8.5 hrs. CB3-BC1 to be paired and crew to be stationed at Bangalore: 17.5 hrs. CB4-BC2 to be paired and crew to be stationed at Chennai: 8.75 hrs. Total layover time = 52 hrs 30 minutes .~) Optimal solution Figure 9.20: Optimal solutions to Example 9.9: IV-II broken ‘The upper bound reduced to 14 from 16 and close to the lower bound 13. Since all the solutions obtained are TSP feasible, no further reduction is possible in the280 Operations Research: Concepts, Problems & Solutions olution to the problem is to follow the tour WH-ITIL-TV-L.Wy with minimum possible distance of 14 units : Summary “The structure of assignment problems. various types of problems, the solution methods for assignment problems. and travelling salesman’s problems and solution methods have been introduced in this chapter. « An assignment problem has the requirement that the number of resources/. facilities and the number of tasks/jobs have to be equal. A square matrix will show the cost of assigning each facility to each task. The objective of ent problems is to minimize the total cost of assigning the tasks to x Lewty There are two constraint sets: row constraints and column constraints: ass the facilities ‘The decision variables are binary in this problem: xjj = {0, 1}; ¥é, 7 If a problem is unbalanced. the number of rows not equal to the number of columns. make it to a balanced problem by adding required number of rows/columns of cost zero. Apply Hungarian assignment algorithm to solve the balanced problem (i) Subtract the minimum cost in each row from the elements of the row and then subtract the minimum cost in each column from the elements of the column (ii) Draw minimum number of horizontal and vertical lines connecting all zeros. If the number of lines required to cover all zeros is the size of the matrix, a feasible solution exists. (iii) If not, subtract the minimum among the uncovered cells from the un- covered cell elements and add it to the cells with horizontal and vertical crossing. (iv) Select the single zero rows or columns and assign the those zeros as the selection. Cross out other zeros on the corresponding row and column. Do this till all the rows and columns get one zero selected. © A maximization assignment problem can be converted to a minimization problem by subtracting all elements from the maximum element and writing the results in the respective cells, # Solve this matrix as a minimization problem and get the optimal solution. Use the maximization values on the selected cells to get the maximum benefit. aed‘Assignment Problems 291 of problems such as crew scheduling, prepare the cost matrix en information and solve as an assignment problem © Travelling salesman’s problem (TSP) has the requirement that a person has to visit all the nods and come back to the starting node covering the minimum. possible distance. Assignment solution becomes the lower bound for the travelling salesman’s problem which may not be TSP feasible. Avoid sub tours by introducing co into the corresponding cells to remove sub-tours and solve again to get the solution. « For variety from the giom most optimal eme of cars from the de- customer locations. What snum total distance that the wld cover to serve the Problem 9.3: Six wagons are available @ stations A. B.C. D. E. and F. These @fe required at stations I. I. IIL IV, Vv, and VL. The distance between different fin kan) is given below. A 20 23 18 16 20 B 350 20 17 Bon c 6 30 40 8 7 D6 7 10 100 9 E i$ 19 28 60 70 F 10 20 40 55 How should the wagons be assigned so that the total distance covered is min- imized? Problem 9.4: Six jobs: A-F have to be processed by six machines available. The average time the jobs take in each of these machines is tabulated below. It is proposed to purchase a new machine to replace one of the existing machines. The estimate of the average time on the new machine is also available. Is it advantageous to purchase this new machine? If yes. which machine should be replaced by this new machine? Average time on machines A 8 8 Bo9 0 8 7 8 9 10 C8 7 8 8 8 6 8 D2 13144 14 15 4 oI E9 9 9 8 8 0 9 Fo7es8 9 9 9 8 8 Problem 9.5; A team of five horses and riders entered the finals of a jumping contest. The number of penalty points that each rider can commit when riding each horse is given below.Assignment Problems Rider Horse RI R2_R3_ Rd RS wHo506306406d7 HD 2 3 7 6 5 HB 4°15 2 4 HH 6 8 1 2 3 Hw 4 2 5 7 4 Determine the most optimam assign- ment of horse to riders that will mini- mize the total penalty points to the team. Problem 9.6: Five men are available to do five different jobs. From the past records, the time (in 2 hrs) that each men takes to do each job is known and given in the following table. Determine the optimum assignment solution. Check if the solution is unique. If not, get all the other solutions. Problem 9.7: Consider the problem of assigning five operators to five ma- chines. The assignment costs are given below: MI | M2] M3 | M4 | M5) Al7|[7/-[418 Blo le[4j/5/|6 Ciujsy7}-s Dio | 4 8 iol 4 E[sj/7[/9)/u/3) Find the optimum cost assignment 293 of machines to operators. Problem 9.8: Solve the following as- signment problem. (AT8y4] 21611] B/0/915\5/4) C/3/819 2/6) Di4/3/1/3/3) [ET915/81915 Problem 9.9: Four drivers are available to be assigned to five routes. The ex- pected cost per trip on each route when each of these drivers are assigned is shown below. Route Divr A B C D £E 62 78 50 101 82 Tl 84 61 73 59 87 92 111 71 81 48 64 87 77 80 Select four routes from the available such that the expected cost will be mini- mum. Which route will be left without operation? 28a Problem 9.10: Determine an optimum cost assignment for the cost matrix given below. Machines Job 1 2 3 4 5 6 A oll 17 8 16 20 15 B 9 7 12 6 15 13 Cc 13 16 15 12 16 8 D 21 24 17 28 26 15 E 4 10 12 11 15 6 Problem 9.11: A shipping company has one excess truck each after preparing the schedules in the cities C1, C2, C3. C4, and C5. At the same time, they have deficit of one truck each in the cities D1, D2, D3, D4. DS. and D6. The cost of Wansportation and handling between the cities with deficit truck and cities with excess truck are as follows.