PRICE-SETTING BASED COMBINATORIAL AUCTION

APPROACH FOR CARRIERS’ COLLABORATION IN LESS

THAN TRUCKLOAD TRANSPORTATION

Bo Dai and Haoxun Chen

Laboratoire d'Optimisation des Systèmes Industriels (LOSI), Institut Charles Delaunay (ICD), UMR CNRS STMR 6279

Université de Technologie de Troyes, 12 rue Marie Curie, BP 2060, 10010, Troyes Cedex, France

Keywords: Collaborative transportation planning, Combinatorial auction, Price-setting method, Lagrangian relaxation.

Abstract: In collaborative logistics, multiple carriers may form an alliance to optimize their transportation operations

through sharing transportation requests and vehicle capacities. In this paper, we study a carriers’

collaboration problem in less than truckload transportation with pickup and delivery requests. After

formulating the problem as a mixed integer programming model, an iterative price-setting based

combinatorial auction approach based on Lagrangian relaxation is proposed. Numerical experiments on

randomly generated instances demonstrate the effectiveness of the approach.

1 INTRODUCTION

In collaborative logistics, multiple carriers may form

an alliance to optimize their transportation

operations by sharing vehicle capacities and delivery

requests. The objective of the collaboration is to

eliminate empty back hauls, to raise vehicle

utilization rate, and thus to increase the profit of

each carrier involved.

In practice, two types of transportation services

are often provided: truckload (TL) transportation

and less than truckload (LTL) transportation. Until

now, most studies on collaborative logistics were

focused on TL transportation (Kwon et al. 2005;

Ergun et al. 2007a, 2007b; Lee et al. 2007), few

papers studied collaborative logistics problems in

LTL transportation (Krajewska and Kopfer, 2006;

Houghtalen et al., 2007; Berger and Bierwirth 2010).

Combinatorial auction has been applied to

truckload transportation service procurement and

carriers’ collaboration. However, previous studies

adopt a quantity-setting based combinatorial auction

approach. In each round, carriers (bidders) submit

prices on various bundles of requests. The

auctioneer then makes a provisional allocation. The

approach requires the pre-selection of preferable

bundles by each carrier and the resolution of a NP-

hard winner determination problem by the

auctioneer in each round. The price setting based

auction approach, to be adopted in this paper, can

overcome the two difficulties.

In this paper, we study a carrier collaboration

problem in less than truckload transportation with

pickup and delivery requests (quoted as CCPLTL

hereafter), where multiple carriers constitute a

transportation alliance for sharing their vehicle

capacities and transportation requests. A price-

setting based iterative combinatorial auction

approach is proposed for reallocating the requests

among the carriers. In the approach, the role of the

auctioneer is to set and update the price of serving

each request, and its objective is to maximize the

total profit of the whole alliance. Each bidder

(carrier) selects its preferable requests to serve by

maximizing its individual profit based on the prices

proposed by the auctioneer. The price update is

based on Lagrangian relaxation. When the auction

process is terminated, if there are still some requests

selected by more than one carrier, the conflict is

resolved by using a random method. The

effectiveness of the approach is demonstrated by

numerical experiments on random generated

instances.

2 PROBLEM DESCRIPTION

In the carriers’ collaboration problem considered,

407

Dai B. and Chen H..

PRICE-SETTING BASED COMBINATORIAL AUCTION APPROACH FOR CARRIERS’ COLLABORATION IN LESS THAN TRUCKLOAD TRANSPOR-

TATION.

DOI: 10.5220/0003186204070413

In Proceedings of the 3rd International Conference on Agents and Artiﬁcial Intelligence (ICAART-2011), pages 407-413

ISBN: 978-989-8425-40-9

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

multiple carriers operating in a transportation

network form a collaborative alliance to share their

transportation requests and vehicle capacities, in

order to increase their vehicle utilization rates and

reduce their empty back hauls. Initially, each carrier

has acquired certain requests from shippers (the

customers of the carrier), where each request is

specified by a pickup location and a time window, a

quantity, and a delivery location and a time window.

We assume that all requests acquired by the alliance

are available to all carriers, this implies that all the

requests must be reallocated among all the carriers

by using a collaborative transportation planning

method. If a request acquired by a carrier is not

served by itself, then the carrier has to transfer part

of the revenue of the request paid by a shipper to the

carrier serving the request. The objective of the

collaborative planning is to find an allocation of

requests to each carrier as well as optimal vehicle

tours for executing the allocated requests subject to

the capacity constraint of each vehicle and the time

window constraints of each request so that the total

profit of the alliance is maximized. After the request

reallocation, the profit of the alliance must be fairly

allocated among all the carriers so that they are

willing to remain in the alliance. For simplicity, in

this study we assume that all carriers have vehicles

of the same capacity and each carrier uses the same

tariffs to calculate its transportation costs.

3 PRICE-SETTING BASED

COMBINATORIAL AUCTION

FRAMEWORK

Motivated by a combinatorial auction mechanism

for truckload transportation service procurement

(Kwon et al., 2005; Lee et al., 2007), we propose a

combinatorial auction framework with associated

models for CCPLTL. In this framework, the

reallocation of transportation requests among

carriers is realized through a multi-round

combinatorial auction. Two types of actors exist in

our framework, auctioneer and bidders. The

auctioneer is a virtual coordinator who sets and

updates the price of serving each request. The

objective of the auctioneer is to maximize the total

profit of the alliance subject to the constraint that

each request is allocated to at most one carrier

finally. The bidders, who are the carriers, select their

preferable requests to serve by maximizing their

individual profits based on the prices proposed by

the auctioneer. Contrary to the quantity-setting based

combinatorial auction which requires the pre-

selection of preferable bundles by each carrier in

each round, our proposed auction is a lagrangian

relaxation based price-setting auction which does

not require the pre-selection. The auction framework

we propose consists of the following steps.

(1) Before the auction, each carrier (bidder)

submits its requests open to auction to the auctioneer

through a common platform. These requests are

recorded in a request pool for auction of the

auctioneer. Every request has an ask price offered by

a shipper (the amount of money paid by the shipper

for the service of the request). This price is kept by

the carrier who receives the request and will not be

known by other carriers.

(2) The auctioneer sets an initial price (referred to

as outsourcing price hereafter) for serving each

request in the pool, which is equal to or less than the

ask price of the request.

(3) The bidders express their selections of requests

based on the current outsourcing prices of all

requests announced by the auctioneer. All requests

in the pool are available to each bidder, and each

bidder selects a set of available requests to serve to

maximize its own profit. The decision problem of

each carrier is referred to as a bidding problem

which will be formulated in the next section.

(4) The auctioneer adjusts the outsourcing price of

each request. By relaxing the constraints that each

request is served by at most one bidder using

lagrangian relaxation, the prices are adjusted based

on the subgradient which is defined as the violations

of the relaxed constraints by the current request

selections of all bidders.

(5) Repeating the above steps (3) and (4) until a

stopping criterion is satisfied, namely, all requests in

the pool are reallocated to at most one carrier or the

current best allocation can not be improved in a

given number of iterations, or a given number of

iterations are achieved.

(6) If there are still some requests selected by more

than one bidder after the above mentioned iterative

auction process is terminated, a random conflict

resolution method is applied to reallocate the

requests to the bidders.

(7) After the iterative auction process, each bidder

gains a profit by serving some requests (referred to

as pre-profit hereafter), which is calculated by

solving the relevant bidding problem. The auctioneer

holds a residual profit which is the difference

between the total profit of the alliance and the pre-

profits of all the carriers. This residual profit is

ICAART 2011 - 3rd International Conference on Agents and Artificial Intelligence

408

redistributed to all bidders based on a profit sharing

mechanism, which ensures that the profit of each

carrier gained with the collaboration is no less than

its profit gained without collaboration. This

mechanism will be discussed in our future work

The interactions between auctioneer and multiple

bidders in each round are illustrated in Figure 1.

Figure 1: The interactions between auctioneer and

multiple bidders in each round.

The advantages of our auction framework are

explained as follows: the ask price of each request is

reserved by the bidder who owns this request and

not known by other carriers; the bidders need not

submit bids in forms of bundles of requests but

select their preferable requests, this can significantly

reduce the computation complexity for considering

exponential number of bundles of requests (

for n

requests). In addition, our framework extends the

work of Kwon et al. (2005) and Lee et al. (2007) for

truckload transportation service procurement to

carrier collaboration in LTL transportation.

4 FORMULATION OF THE

COMBINATORIAL AUCTION

PROCESS

In this section, we formulate the combinatorial

auction process for CCPLTL, which includes a

global optimization model for the alliance, the

bidding problem for each carrier, and the iterative

price adjustment by auctioneer.

4.1 Global Optimization Model

For simplicity, we assume that no more than one

transportation request is associated with each node

in the transportation network considered.

Indices

i, j, m = 1, . . . , N node index, where N represents

the number of nodes in the transportation network.

The nodes include the locations of all shippers, the

locations of all customers of the shippers, and the

vehicle depots of all carriers.

k = 1, . . . , K carrier index, where K represents the

number of carriers.

l = 1, . . . , L request index, where L represents the

number of requests.

Parameters

the set of requests whose pickup site is node i

the set of requests whose delivery site is node i

quantity delivered on request l

price paid by a shipper to serve request l

C vehicle capacity

the number of vehicles owned by carrier k

the depot of carrier k

shipping cost from node i to j for each vehicle

where c

ij =

and the triangle inequality c

im +

≥ c

holds for any i, j, m with

,mim j

travelling time of a vehicle from node i to node j

the earliest service time at node i

the latest service time at node i

a large number, T

= b

 a

Variables

q quantity transported through arc (i, j) by carrier k

the number of times that arc (i, j) is visited by

vehicles of carrier k

y 1 if request l is served by carrier k; otherwise 0

t the time at which a vehicle of carrier k leaves

node i

With the notation, the total profit optimization

problem of the alliance can be formulated as a mixed

integer programming model P as follows:

Model P:

11 111,

KL KN N

llk ijij

kl kijji

Max p y c x

 

 









  

(1)

Subject to:

1, 1,

,1,...,, ,{ } 1,..., ,

ij ji k

jji jji

xxiNiok K

 





(2)

, , 1,..., , , 1,..., ,

ij ij

Cij Nijqx k K   

(3)

1, 1,

1,..., ,

ij ji l lk l lk

j ji j ji lP lD

qqdydy

   

 

















 

(4)

1, 1,

,1,...,,

jjo jjo

oj jo

xxkK

 





(5)

, 1,..., ,

jjo

Wk K







(6)

PRICE-SETTING BASED COMBINATORIAL AUCTION APPROACH FOR CARRIERS' COLLABORATION IN

LESS THAN TRUCKLOAD TRANSPORTATION

409

, 1,..., ,1







(7)

, 1,..., , ,1{}1,...,,

ij k

jji

xi Niok K







(8)





, , 1,..., , , , ,11,...,,

kk k k

j i ij ij ij ij

ij Nij i jtttxT x o k K  

(9)

, 1,..., ,

ko o

oj l lk l lk

jjo lP lD

qdydykK

  



 

(10)

, , 1,..., , , 1,..., ,0,

ij ij

ij Ni jk Kxx 

(11)





1,..., , 1,..., ,0,1 , ,

lk lk

yyZlLkK

(12)





1,..., , 1,..., ,0,1 , ,

lk lk

yyZlLkK

(13)

1,..., , , 1,..., ,0,

ii i

iNikKat b o

(14)

The objective function (1) represents the total profit

of the alliance. Constraints (2) ensure that the

number of vehicles leaving from each node is equal

to the number of vehicles arriving at this node.

Constraints (3) are the vehicle capacity constraints.

Constraints (4) are the flow conservation equations,

assuring the flow balance at each node. Constraints

(5) guarantee that the number of vehicles of carrier k

leaves from depot o

is equal to the number of

vehicles arrives at this depot. Constraints (6) imply

that no more than W

vehicles can be used by carrier

k. Constraints (7) guarantees that each request is

allocated to at most one carrier. Constraints (8)

ensure that each pickup/delivery node is visited by

each carrier at most once. Constraints (9) denote the

relations between departure times

t . Constraints

(10) imply that only empty vehicle is returned to the

depot of each carrier. Constraints (14) are time

window constraints for pickup and delivery

operations of the requests. Since at most one request

is associated with each node, the constraints are

associated with the nodes.

Note that a solution of model P does not

completely define vehicle tours for each carrier, but

they can be constructed based on the solution by

applying a tour generation method proposed in our

previous work (Dai and Chen, 2009a, 2009b).

4.2 Iterative Auction based on

Lagrangian Relaxation

Based on model P, we propose an iterative price-

setting combinatorial auction for the total profit

maximization of the alliance. The prices in this

auction are the Lagrange multipliers we introduce

for relaxing constraints (7) in the model. The relaxed

problem can be decomposed into several

subproblems, one for each carrier, which determines

the preferable requests to bid by the carrier at the

current price for serving each request given by the

auctioneer. This subproblem is referred to as the

bidding problem of the carrier. The price adjustment

of the auctioneer in each iteration (round) is based

on the subgradient defined as the violations of the

relaxed constraints by the current request selections

of all carriers. The auction process will be

terminated until some condition is satisfied.

4.3 Bidding Problem for each Carrier

To formulate the bidding problem for each carrier,

we rewrite objective function (1) as objective

function (15),

11 111,

KL KN N

llk ijij

kl kijji

Min p y c x

 

 









  

(15)

which transforms model P into an equivalent

minimization problem. We then relax constraint (7)

by introducing the Lagrange multipliers





,1,...,,0



 

, leading to the following

relaxed problem SP

11 111,

KL KN N

llk ijij

kl kijji

KL KN N

llk ijij

kl kijji

py cx

ZMin

py cx

Max





 



 



























 











 





  



  





(16)

subject to constraints (2) to (6), (8) to (14).

The relaxed problem can be decomposed into k

subproblems

, one for each carrier k, which is

the bidding problem of carrier k.

Subproblem



111, 1

111,

LNN L

R l lk ij ij l lk

lijjil

LNN

l l lk ij ij

lijji

ZMax

Max

py cx y

py cx











  







 





 



(17)

subject to constraints (2) to (6), (8) to (14)

associated with carrier k.

Accordingly, an upper bound for model P can be

calculated by (18).

RLRl









(18)

In model SP

, p

is the ask price of request l (the

price paid by a shipper to serve request l), and

Lagrange multipliers λ

are given by the auctioneer.

ICAART 2011 - 3rd International Conference on Agents and Artificial Intelligence

410

The combinatorial auction process considered then

consists of the following steps.

Step1, auctioneer sets an outsourcing price for each

request, i.e., the value of (p

-λ

Step2, Each bidder (carrier) determines which

requests to serve by solving its bidding problem

Step3, the auctioneer adjusts the outsourcing prices

by updating λ

based on the violations of the relaxed

constraints.

Step4, repeat the above process until no constraint is

violated, i.e., each request is allocated to at most one

carrier, or a limit number of iterations is achieved.

For the latter case, the allocation is infeasible

because some requests are allocated to more than

one carrier. In this case, the auctioneer randomly

allocates each of the requests to one carrier.

4.4 Iterative Price Adjustment by

Auctioneer

As we mentioned, the subgradient method (Fisher,

2004) is used to update the Lagrange multipliers.

Given an initial value



, the value of



in the m-th

iteration of the auction, denoted by



, is calculated

by the equation (19).

1,0 1,..., ,

llm

ax t l L

























(19)

In equation (19), y

is provided by carrier k after

solving its bidding problem

at the m-th iteration;

is a positive scalar step size, it is set to a fixed

amount



initially. If the objective value of SP

not improved in a given number of iterations,



halved, i.e.,





Three stopping conditions are used in the

combinatorial auction iteration process. If one of

them is satisfied, the process will be terminated.

(1) No constraint of (7) in model P is violated.

(2) The number of iterations performed exceeds a

predefined number.

(3) The current step size



is smaller than a given

small value.

5 NUMERIC EXPERIMENTS

Up to now, there are no benchmark instances for

CCPLTL, so all instances used to evaluate the

performance of our proposed combinatorial auction

approach are generated randomly. These instances

are grouped into two sets, which are different in the

generation of the coordinates of each node.

All instances involve three carriers (K = 3)

whose vehicles have the same capacity C = 10. The

number of transportation requests L is set to (N-K)/2

(L = 15). The requests are generated by randomly

choosing a pickup node i (i ≠ any depot) and a

delivery node j (j ≠ i and any depot), each node is

associated with at most one request; each request is

associated with a randomly generated quantity of

freight d which is no larger than a predefined

number, which is set to 2, 5, 10 for generating 5, 5, 5

instances respectively in each set. Given a request

with pickup node i and delivery node j acquired by a

carrier from its shipper, the ask price of serving the

request is set as α*(1+β)*d*(c

)/C, where o

and β denote the depot of the carrier and the profit

margin (set to 0.05) of the carrier, respectively, d is

the quantity of the request, (c

) is the direct

shipping cost for the request, α is the vehicle

utilization rate of the carrier. The time windows are

generated in the following way: the time interval for

serving all requests is set to [0, 144] (1440 minutes =

24 hours, time unit is taken as 10 minutes); the

earliest service time a

at pickup node i is randomly

chosen from 0 to 60, the latest service time b

randomly chosen from (a

+ 6) to 72; the earliest

service time a

at delivery node j is randomly chosen

from 72 to 132, and latest service time b

randomly chosen from (a

+ 6) to 144. The bidding

problem of each carrier is solved by the MIP solver

of ILOG Cplex 11.2 for all instances. The time limit

for the resolution of each bidding problem is set to

one hour. The initial value of each Lagrange

multiplier is set to 0; the step size



is initially set to

50 and its minimum value is to 0.001;



is halved if

the objective value of SP

is not improved in 10

iterations; the maximum number of iterations for the

auction is set to 200.

The number of nodes is set to 33. For each

instance in the first set, the coordinates data of the

first 33 nodes of benchmark instance R101 of

VRPTW (Solomon, 2005) are used, where node 5,

17, 11 are chosen as the depots of the three carriers

respectively. Each carrier has 10 vehicles. For each

instance in the second set, the coordinates of each

node are randomly generated from 6666 square.

After the coordinates of all nodes are generated, the

Euclidean distance d

between any two nodes i and j

is calculated. Without loss of generality, we set c

= c

. Every carrier randomly selects a node as its

depot which is different from the depot nodes of all

other carriers; each carrier owns a number of

vehicles randomly chosen from 1 and 10.

PRICE-SETTING BASED COMBINATORIAL AUCTION APPROACH FOR CARRIERS' COLLABORATION IN

LESS THAN TRUCKLOAD TRANSPORTATION

411

All computational results are given in Table 1

and 2, where the row Quantity ≤ 2, ≤ 5, ≤ 10

indicates the maximal pickup/delivery quantity

generated for each request. The results presented in

row Opt are obtained by solving the global

optimization model P of each instance using Cplex

11.2 with a time limit of 2 hours. The row CAFLB

and CAFUB denote the lower bound and the upper

bound obtained by our combinatorial auction

approach, respectively. The row Gap denotes the

percentage difference between CAFLB and CAFUB.

The row Iteration and Time denote the number of

iterations and the time (in seconds) for solving each

instance by our combinatorial auction approach.

Table 1: Results for the first set of fifteen instances.

Quantity

≤ 2

1 2 3 4 5

Opt

Time (s)

2431.5 2516.2 2602.6 2348.1 2279.2

141 690 402 101 116

CAFLB

CAFUB

Gap (%)

Iteration

Time (s)

2431.5 2516.2 2426.4 2348.1 2279.2

2431.5 2516.2 2616.23 2348.1 2279.2

0 0 7.8 0 0

58 90 81 50 66

450 1278 1423 514 936

Quantity

≤ 5

6 7 8 9 10

Opt

Time (s)

2618.7 2536.9 2499.7 2398.2 2426.9

275 7200 836 1858 1672

CAFLB

CAFUB

Gap (%)

Iteration

Time (s)

2618.7 2536.9 2499.7 2398.2 2426.9

0 0 0 0 0

81 72 45 73 152

1761 41772 7850 9006 7061

Quantity

≤ 10

11 12 13 14 15

Opt

Time (s)

2138.2 2173.7 1947.6 2211.7 2504.7

961 7200 590 7200 7200

CAFLB

CAFUB

Gap (%)

Iteration

Time (s)

2138.2 2173.7 1947.6 2211.7 2504.7

0 0 0 0 0

81 68 49 75 54

3891 32915 1911 44702 42243

From the above two tables, we can see that our

combinatorial auction approach can find a globally

optimal solution for most instances except for

instances no. 3, no. 22, and no. 25. For the three

instances, our approach can find a fairly good

solution.

Table 2: Results for the second set of fifteen instances.

Quantity

≤ 2

16 17 18 19 20

Opt

Time (s)

502.9 965.9 881 1314.4 514.9

2308 292 44 62 1000

CAFLB

CAFUB

Gap (%)

Iteration

Time (s)

502.9 965.9 881 1314.4 514.9

0 0 0 0 0

49 82 85 58 83

10266 4098 697 338 2246

Quantity

≤ 5

21 22 23 24 25

Opt

Time (s)

2445.4 816.8 380.1 690.2 1233.6

776 128 2250 3914 705

CAFLB

CAFUB

Gap (%)

Iteration

Time (s)

2445.4 707.7 380.1 690.2 1082

2445.4 838.701 380.1 690.2 1266.9

0 18.5 0 0 17.1

52 115 62 54 56

9376 2696 17232 35113 992

Quantity

≤ 10

26 27 28 29 30

Opt

Time (s)

651 746.5 95.2 706.1 1062.4

5936 406 110 291 6359

CAFLB

CAFUB

Gap (%)

Iteration

Time (s)

651 746.5 95.2 706.1 1062.4

0 0 0 0 0

57 46 41 81 57

84407 13805 947 3960 50037

6 CONCLUSIONS

The carriers’ collaboration problem in less than

truckload transportation with pickup and delivery

requests has been studied in this paper. A

Lagrangian relaxation based price-setting

combinatorial auction approach is proposed for the

total profit maximisation of the alliance in the

collaboration. Numerical experiments on thirty

randomly generated instances demonstrate the

effectiveness of the approach. The main advantage

of our approach is that the decision marking of each

carrier is made in an autonomous and decentralised

way, there is no confidential information of a carrier

revealed to other carriers, so the approach is more

implementable than a centralised approach where all

information is shared among the carriers. In our

future work, we will design a fair profit allocation

mechanism to keep the persistence of the

collaborative alliance.

ICAART 2011 - 3rd International Conference on Agents and Artificial Intelligence

412

REFERENCES

Berger, S., Bierwirth, C. (2010) Solutions to the request

reassignment problem in collaborative carrier

networks, Transportation Research Part E, Vol. 46,

No. 5, pp.627-638.

Dai, B., Chen, H. X., (2009a) Mathematical model and

solution approach for collaborative logistics in less

than truckload (LTL) transportation, 39th

International Conference on Computers & Industrial

Engineering, Troyes, France.

Dai, B., Chen, H. X., (2009b) A benders decomposition

approach for collaborative logistics planning with LTL

transportation, The Third Int.l Conf. on Operations

and Supply Chain Management, Wuhan, China.

Ergun, Ö., Kuyzu, G., Savelsbergh, M. (2007a) Shipper

collaboration, Computers & Operations Research,

Vol. 34, No. 6, pp.1551-1560.

Ergun, Ö., Kuyzu, G., Savelsbergh, M. (2007b) Reducing

truckload transportation costs through collaboration,

Transportation Science, Vol. 41, No. 2, pp.206-221.

Fisher, M. L. (2004) The Lagrangian relaxation method

for solving integer programming problems,

Management Science, Vol. 50, No. 12 Supplement,

pp.1861–1871.

Houghtalen, L., Ergun, Ö., Sokol, J. (2007) Designing

allocation mechanisms for carrier alliances, available

athttp://www.agifors.org/award/submissions2007/

Houghtalen_paper.pdf

Krajewska, M. A., Kopfer, H. (2006) Collaborating freight

forwarding enterprises: request allocation and profit

sharing, OR Spectrum, Vol. 28, No. 2, pp.301-317.

Kwon, R. H., Lee, C. G., Ma Z. (2005) An integrated

combinatorial auction mechanism for truckload

transportation procurement, Working paper, available

at http://www.mie.utoronto.ca/labs/ilr/MultiRound.pdf

Lee, C. G., Kwon, R. H., Ma Z. (2007) A carrier’s optimal

bid generation problem in combinatorial auctions for

transportation procurement, Transportation Research

Part E, Vol. 43, No. 2, pp.173–191.

Solomon, M.M (2005) VRPTW Benchmark Problems,

available at http://w.cba.neu.edu/~msolomon/r101.htm

PRICE-SETTING BASED COMBINATORIAL AUCTION APPROACH FOR CARRIERS' COLLABORATION IN

LESS THAN TRUCKLOAD TRANSPORTATION

413