An evaluation methodology for city logistics

This article was downloaded by: [Radboud Universiteit Nijmegen] On: 17 April 2015, At: 07:37 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Transport Reviews: A Transnational Transdisciplinary Journal Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ttrv20

An evaluation methodology for city logistics Eiichi Taniguchi & Rob E.C.M. Van Der Heijden Published online: 26 Nov 2010.

To cite this article: Eiichi Taniguchi & Rob E.C.M. Van Der Heijden (2000) An evaluation methodology for city logistics, Transport Reviews: A Transnational Transdisciplinary Journal, 20:1, 65-90, DOI: 10.1080/014416400295347 To link to this article: http://dx.doi.org/10.1080/014416400295347

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution,

Downloaded by [Radboud Universiteit Nijmegen] at 07:37 17 April 2015

reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-andconditions

TRANSPORT REVIEWS, 2000, VOL. 20, NO. 1, 65 ± 90

An evaluation methodology for city logistics² EIICHI TANIGUCHI³ Department of Civil Engineering, Kyoto University, Yoshidahomachi, Sakyo-ku, Kyoto 606-8501, Japan


and ROB E. C. M.

VAN DER

HEIJDEN

Faculty of Systems Engineering, Policy Analysis and Management, Delft University of Technology, Jafalaan 5, PO Box 5015, 2600GA, Delft, The Netherlands (Received 12 November 1998; accepted 22 January 1999 ) This paper presents a methodology for evaluating city logistics initiatives using a dynamic tra c simulation with optimal routing and scheduling. This methodology was applied to a test road network. The performance of three city logistics initiatives, advanced routing and scheduling systems, cooperative freight transport systems and load factor controls were assessed in terms of total costs and CO2 emissions by pickup/delivery trucks operations within the network. Results indicated that these initiatives were not only eŒective for reducing total costs, but also for CO2 emissions. The methodology presented here allows city planners to quantitatively evaluate city logistics initiatives.

1. Introduction 1.1. City logistics initiatives Urban freight transport has become an important component of urban planning. The rationalization of urban freight transport is essential for sustainable economic growth. However, there are now many problems to overcome such as tra c congestion, environment and energy conservation. Freight carriers are expected to provide higher levels of service within the framework of Just-in-Time transport systems with lower costs. To help solve these di cult problems, several city logistics initiatives have been proposed, including: 1. 2. 3. 4. 5.

Advanced information systems. Cooperative freight transport systems. Public logistics terminals. Load factor controls. Underground freight transport systems.

1.2. Advanced information systems Advanced information systems have become important in rationalizing existing logistics operations. In general, advanced information systems for pickup/delivery trucks operations have three important functions: ² This article continues the Transport Reviews series in memory of Jim Cooper. [Editor] ³ e-mail: taniguchi@ urbanfac.kuciv.kyoto-u.ac.jp Transport Reviews ISSN 0144-1647 print/ ISSN 1464-5327 online Ó 2000 Taylor & Francis Ltd http://www.tandf.co.uk / journals/tf /01441647.html

66

E. Taniguchi and R. E. C. M. van der Heijden


1. 2. 3.

To allow drivers and the control centre to communicate with each other. To provide the real time information on the tra c conditions. To store detailed historical pickup/delivery trucks operations data.

The third function has not been fully discussed in the literature, but it is very important for rationalizing logistics operations. One successful application of historical operations data was experienced by a Japanese milk-producing company. After introducing a satellite-based information system for 1 year, the company reduced the number of pickup/delivery trucks by 13.5% (from 37 to 32 vehicles ) and increased their average load factor by 10% (from 60 to 70% ). A computer-base d system was used to store detailed historical data of the pickup/delivery trucks operations, including times of starting/arriving times at the depot and customers as well as the waiting times, travelling speeds and routes travelled. The company analysed these data and changed their routes and schedules substantially to increase the e ciency of their vehicle ¯ eet. This type of system can reduce both freight transport and environmental costs within a city. 1.3. Cooperative freight transport systems Several researchers have investigated cooperative freight transport systems (Ruske 1994, Taniguchi et al. 1995, Kohler 1997 ) that allow a reduced number of trucks to be used for collecting or delivering the same amount of goods. Based on a survey by Kohler (1997 ), it is remarkable to see competitive freight carriers cooperating in delivering goods to the inner city of Kassel in Germany. A neutral freight carrier collects goods from ® ve freight carriers and delivers them to shops in the inner city. After introducing this system the total time travelled by trucks was reduced and queues of trucks for waiting to deliver goods on streets was also reduced. Originally this system started with 10 freight carriers, and now ® ve companies remain in the cooperative system. Another outstanding case is the cooperative delivering system among 11 department stores in Osaka, Japan. In this system, basically two department stores having depots adjacent each other exchange their goods to be delivered in the neighbourhood of the depot. This led to the considerable reduction of travel time for trucks, person work-hours and total costs. As observed in these cases, cooperative freight systems can substantially reduce transport costs as well as environmental impacts. 1.4. Public logistics terminals Public logistics terminals in areas surrounding a city can be helpful in promoting cooperative freight transport systems (Janssen et al. 1991, Duin 1997, Taniguchi et al. 1997 ). A good example of this platform for city distribution can be seen in Monaco. This platform is provided by the government and operated by a private freight carrier for delivering goods to city areas. This company is subsidized by the government to provide a delivery service with cheaper prices than normal. This system helps reduce the required number of trucks used for deliveries. In Japan, the ® rst multi-functional logistics terminal is to be built in Seki near Nagoya. This logistics terminal is referred to as a logistics town and has various functions such as the trans-shipment of goods, assembling products during distribution, warehouses and wholesale markets. This project is being planned and executed by a group of companies from various kinds of industries with the support of the national, prefecture and municipal governments.


City logisitics evaluation methodology

67

1.5. Controlling load factors Controlling the loadings of pickup/delivery trucks is a relatively new initiative compared with conventional regulations such as vehicle weight limits, designated times for trucks to enter the city centres and the control of vehicle emissions. Two European cities (Copenhagen and Amsterdam ) have introduced a certi® cate system for freight carriers who deliver or collect goods within the central city areas in 1998. In Copenhagen, only vehicles with a certi® cate (green sticker ) are allowed to use public loading /unloading terminals in the inner city. This certi® cate can only be issued to vehicles satisfying the following conditions: (1 ) load factor is > 60 % ; and (2 ) vehicle is < 8 years old. Companies owning vehicles are required to produce a report on the load factors of their vehicles every month. To maintain certi® cation, a company must have an average load factor during the previous month > 60% . In Amsterdam vehicles weighing > 7.5 tons are not permitted to go through streets other than main ones. However, vehicles > 7.5 tons can obtain a special certi® cate to enter these streets, if they satisfy the following conditions: (1 ) load factor is > 80% ; (2 ) length < 9 m; and (3 ) engine must satisfy Euro II emission standards. The police inspect the load factor of speci® c vehicles on the road. This initiative assumes that higher load factors produce lower the environmental impacts. 1.6. Undergroun d freight transport systems Underground freight transport systems are innovative solutions for urban freight transport problems. Koshi et al. (1992 ) estimated the impacts of building an underground freight transport system in the central area of Tokyo, Japan. The results of this study indicate that NOx and CO2 emissions would be reduced by 10 and 18% respectively, that energy consumption would be reduced by 18% and that average travel speed would be increased by 24% . Oishi (1996 ) studied the economic feasibility of the underground freight transport system in Tokyo and concluded that this project has an internal income rate of 10% when the infrastructure is constructed by the public sector. The Dual Mode Truck (DMT ) was developed and tested by the Public Works Research Institute of the Ministry of Construction, Japan. This new type of automated electric truck can travel through an exclusive guided lane in an underground tunnel with an external supply of electricity and can also travel on normal streets operated by a driver with batteries. In The Netherlands a similar idea was proposed (Visser 1997, Duin 1998 ) and the feasibility of underground freight transport system between Aalsmeer and Schiphol Airport for carrying ¯ owers was investigated. An automated guided truck named the Combi-road system was also developed and tested by a group of private companies. 1.7. Private and public partnershi p City logistics initiatives are usually operated by private companies with varying degrees of support provided by the public sector. To realise the full potential of city logistics initiatives, it is, therefore, crucial that an eŒective partnership between both the private and public sector be developed and maintained. 1.8. City logistics evaluation criteria This paper focuses on the evaluation of three of the ® ve city logistics initiatives described above, advanced information systems, cooperative freight transport systems and the control of load factors. These evaluations were undertaken using a dynamic tra c simulation model developed by Taniguchi et al. (1998 ). Here, the

68



emphasis is on evaluating the city logistics initiatives from an environmental point of view. The criteria used in this paper for evaluating each city logistics initiative relate to measuring the eŒects of the initiatives on congestion as well as the environmental impacts. The total travel time and the CO2 emissions produced by pickup/delivery trucks are the primary indicators used in the evaluations. Other indicators could be considered, including NOx emissions, noise, ground vibrations, but it requires further investigations for evaluating them. 1.9. Components of freight transport In general, there are three important components relating to freight transport, economic growth, demand for freight transport and impacts on congestion and environment. These three components are closely related to each other (® gure 1 ). However, it is desirable to change the nature of the relationships between these components and achieve economic growth with less impacts on both congestion and the environment. This means that the demand for freight transport should not substantially increase with economic growth and that congestion and environmental impacts should not increase with the rising demand for freight transport. Hopefully, these two components would even be reduced as the economy grows. Figure 2 shows the relationship between the Gross National Product (GNP ) and freight transport in terms of ton-km in Japan from 1970 to 1994. This ® gure indicates that freight transport in terms of ton-km generally increased as GNP increased. However, after the oil crises in 1973 and 1979 freight transporte d decreased, while the GNP continued to increase. A detailed discussion of this relationship phenomenon is not presented here. City logistics initiatives are expected to change the general positive relationship between the demand for freight transport and the impacts on congestion and environment. Therefore, this paper focuses on analysing this relationship. 1.10. Pickup/delivery truck routing and scheduling This paper focuses on investigating pickup/delivery truck routing and scheduling operations in an urban area where some freight carriers are assumed to have introduced advanced routing and scheduling procedures as well as established a cooperative freight transport system. Moreover, the municipality

Figure 1.

Three components relating to freight transport.



Figure 2.

69

Relationship between freight transport in terms of ton-km and gross national product in Japan (1970-94 ).

regulates the load factors of pickup/delivery trucks. The eŒects of these initiatives on the relationship between the CO2 emissions and the demand for freight transport were predicted. When investigating these eŒects the designated time for pickup and delivery plays an important role. Recently urban pickup/ delivery trucks were required to arrive at their customers within a designated time. A recent survey in Osaka and Kobe in Japan found that freight carriers were required operate with designated arrival times or time windows for 52% of goods delivered and for 45% of goods collected in terms of weight. Vehicle routing problems with time windows (VRPTW ) have been investigated by a number of operations researchers (e.g. Solomon 1987, Koskosidis et al. 1992, Russell 1995, Bramel et al. 1996 ). However, most of this research has been conducted within the framework of a company’s business logistics. The impacts of the behaviour of shippers and freight carriers on the general tra c conditions of the road network have not yet been investigated. However, these impacts are considered to be very important for city planners to evaluate transport policies for alleviating congestion, environmental and energy problems. This paper describes the application of a model for representing the behaviour of urban pickup/delivery trucks as well as passenger cars on a road network. Since freight transport is generally undertaken by private companies in a free market, too much control and regulation by the public sector is not welcomed. The model allowed the eŒects of three city logistics initiatives, advanced routing and scheduling systems based on advanced information systems, cooperative freight transport systems and control of load factors to be investigated. Changes in CO2 emissions within an urban area were estimated.

70

2. Model 2.1. Framework presents a framework of the model applied. The model is two submodels, a model for vehicle (pickup/ delivery truck ) scheduling problem with time windows (VRPTW ) for each well as a dynamic tra c simulation model for the ¯ eet of


Figure 3 composed of routing and company as


Figure 3.

Model framework.



71

pickup /delivery trucks and passenger cars on the road network within the city. The model for VRPTW is de® ned as follows. A depot and a number of customers are de® ned for each freight carrier. A ¯ eet of identical vehicles collects goods from customers and deliver them to the depot or deliver goods to customers from the depot. For each customer a designated time window, specifying the desired time to be visited is also speci® ed. For example, in the case of collecting goods, vehicles depart from the depot and visit a subset of customers to pick up goods in sequence and return to the depot to unload them. A vehicle is allowed to make multiple traverses per day. Each customer must be assigned to exactly one route of a vehicle and all the goods from each customer must be loaded on the vehicle at the same time. The total weight of the goods for a route must not exceed the capacity of the vehicle. The problem is to determine the optimal assignment of vehicles to customers and the departure time as well as the order of visiting customers for a freight carrier. VRPTW explicitly incorporate the departure time of vehicles as a variable to be determined. The optimal assignment of vehicles to customers and the departure time as well as the visiting order of customers for each freight carrier, becomes input to the dynamic tra c simulation model. The dynamic tra c simulation model is based on a macroscopic dynamic simulation BOX model (Fujii et al. 1994 ). This model estimates the average travel time on each link in 30-min intervals. The VRPTW model is then resolved using the updated average travel times on each link obtained from the BOX model. Thus, the average travel times for each link are represented by a step function, in 30-min intervals. The model, therefore, incorporates time-dependent travel times. Successive iterations of both the VRPTW model and the BOX model continues until a pre-de® ned convergence criterion is satis® ed. 2.2. VRPTW Model This section de® nes the mathematical model used to represent the VRPTW that was introduced above. The model minimizes the total cost of distributing goods with truck capacity and designated time constraints. The total cost is composed of three cost components: (1 ) ® xed costs of vehicles; (2 ) vehicle operating costs, that are proportional to the time travelled and spent waiting at customers; and (3 ) delay penalty costs for designated pickup /delivery time at customers. We de® ne the following notations. C ( t0 , X ) t0 t0 X X xl xl n(i ) Nl m cf,l

total cost (yen ) departure time vector of vehicle l at the depot = tl,0 | l 5 l, m assignment and order of visiting customers vector for all vehicles = xl | l 5 l, m assignment and order of visiting customers for vehicle l = n( i) | i 5 l, Nl node number of i -th customer visited by a vehicle total number of customers visited by vehicle l maximum number of vehicles available ® xed cost for vehicle l (yen/vehicle )

72



d l ( xl )

=

1 if vehicle l is used 0 otherwise

ct,l operating cost for vehicle l (yen/min ) Tl tl,0 , xl operating time of vehicle l (min ) cd delay penalty cost at customer (yen/min ) t al,n ( i ) ( tl,0 , x l ) arrival time at node n(i ) of vehicle l that departed from the depot at time tl,0 ten( i) end of time window at customer n(i ) (see ® gure 4 ) tl,n( i) departure time of vehicle l at customer n (i ) d( t, n( i) , ( i 1 1)) minimum travel time at time t from customer n(i ) to customer n(i+ 1 ) tsn( i) start of time window at customer n(i ) (see ® gure 4 ) tc,n( i) loading /unloading time at customer n(i ) W l ( xl ) load of vehicle l (kg ) Wc,l capacity of vehicle l (kg ) The model can be formulated as follows: min C( t0 , X) 5

m

cf,l . d l ( xl ) 1

l5 1

m

ct,l . Tl tl,0, xl ) 1

l5 1 m

cd max 0, tal,n( i) tl,0 , xl 2

ten( i) }

( 1)

l5 1

where Tl ( tl,0 , xl ) 5

Nl

max tl,n( i) 1

d( tl,n( i) , n( i) , n( i 1 1)) , tsn( i1

1)

1 tc,n( i1

1)

2 tl,n( i)

( 2)

i5 0

Subject to Wl ( xl )

£ Wc,l

( 3)

The problem speci® ed by equations (1 ) ± (3 ) is to determine the variable vector X, that is, the assignment of vehicles and the visiting order to customers and the variable t0 , the departure time of vehicles from the depot. Note that n(0 ) and n(Nl+ 1 ) represent the depot in equation (2 ). Figure 4 shows the cost function for vehicle arrivals at customers. The time period ( ten( i) 2 tsn( i) ) de® nes the width of the soft time window. If a vehicle arrives at a customer earlier than tsn( i) , it must wait until the start of the designated time window and a cost is incurred for waiting. If a vehicle is delayed, it must pay a penalty proportional to the amount of time it was delayed. This type of penalty is typically observed in goods distribution to shops and supermarkets in urban areas. The problem described herewith is a NP-hard combinatorial optimization problem. It requires heuristic methods to e ciently obtain an optimal solution. Recently several researchers have applied heuristic algorithms such as genetic algorithms (GA ) (e.g. Thangiah et al. 1991 ), simulated annealing (SA ) (e.g. Kokubugata et al. 1997 ) and tabu search (TS ) (e.g. Potvin et al. 1996 ) to obtain approximate solutions for the VRPTW. Gendreau et al. (1997 ) reviewed the



Figure 4.

73

Cost function for soft time windows.

application of such modern heuristic approaches to VRP and described the potential of such methods for tackling complex, di cult combinatorial optimization problems. The model described in this paper uses a GA to solve the VRPTW. GA was selected because it is a heuristic procedure that can simultaneously determine the departure time and the assignment of vehicles as well as the visiting order of customers. GA generally starts with an initial population of individuals (solutions ) and from these a next generation is produced. Parents of subsequent generations are selected on the basis of their performance or ® tness. Using the parents characteristics, a number of operations are performed (crossover and mutation ) to produce successive generations and to avoid local optimal solutions. Generations are continued to be produced until a satisfactory solution is found. 2.3. Dynamic simulation model The dynamic tra c simulation model is based on a BOX model that was originally developed by Fujii et al. (1994 ). The BOX model is essentially a macroscopic model but because the origin and destination of each vehicle is de® ned, it is actually a hybrid macroscopic/microscopic model. Vehicles are assumed to choose the shortest path when they arrive at a node using an estimated travel time. The BOX model consists of two components, a ¯ ow simulation and a route choice simulation as shown in ® gure 5. A sequence of boxes is used to represent each link. Groups of vehicles ¯ owing out of a box and into the next box during the scanning interval represent the ¯ ow on links. There are two assumptions for modelling links, that is the maximum ¯ ow during a scanning interval is the same for all sections on links and no in¯ ow and out¯ ow is allowed in the middle of links. A consequence of the ® rst assumption, is that only the lowest section of a link can be a bottleneck, where a congestion queue starts. Two states of ¯ ow; congested ¯ ow and free ¯ ow are represented. The time for a vehicle to proceed through a congested queue Tc is given by


74


Figure 5.

Structure of the BOX model.

Tc

5

Fc Ce

( 4)

where Fc number of vehicles in a congestion queue Ce e‚ uent tra c volume The e‚ uent tra c volume is the tra c volume that can ¯ ow out of the lowest section of a link into a lower link. The time that is required to go through the running area without any queue Tf is estimated by Tf 5 Tf 5

Lf ; Vf

if K £

K0 5

Lf . K ; if K > K0 5 Qmax

Qmax Vf Q max Vf

( 5)

( 6)



where

Lf

length of flowing area without any queue

Vf

free running speed

K

traffic density

K0

critical traffic density

Qmax

maximum traffic density

75

The modi® ed BOX model shown in ® gure 6 explicitly describes the ¯ ow of pickup/ delivery trucks that depart from a depot and return to the same depot. Pickup/ delivery trucks are converted to passenger car units and the ® rst-in-® rst-out rule is assumed on all links. The model was further modi® ed to identify the arrival of speci® c vehicles at assigned nodes (customers ). The simulation model described above estimates travel times on each link and allows link costs to be determined. Drivers are assumed to compose a cognitive map for each link based on its estimated link cost. Drivers then choose routes based on their minimum travel cost from the current node to the destination using their cognitive map. It is assumed that all drivers have some experience in driving within the de® ned network. The function for estimating the link cost is: Ck 5 Tkt 1 g k ( 7) where Ck estimated cost on link k Tkt

travel time on link k at time t

g k

disturbance term

In this study the disturbance term g k is assumed to be normally distributed with zero mean and variance r g 2 as represented by g k

Figure 6.

~

N( 0, r g 2 )

Link representation using the modi® ed BOX model.

( 8)

76


Evaluating city logistics initiatives on test network 3.1. Test conditions The model described in the previous section was applied to a test network with 25 nodes and 40 links (® gure 7 ). This network includes three types of roads, urban expressways, arterials and streets with free running speeds of 60, 40 and 20 km/h respectively. Although this network is a hypothetical one, it is similar to Kobe City in Japan. Therefore, the travel times vary for the same distance depending on the type of road used. Note, that the length of links shown in ® gure 5 do not precisely indicate their geometric distance. Any node within the network can both generate and attract passenger car tra c. These nodes are referred to as centroids and are also candidate nodes to be visited by pickup/delivery trucks. Ten freight carriers are


3.

Figure 7.

Test network.

77



assumed to operate a maximum of 12 pickup/delivery trucks in this network. Each freight carrier has one depot whose location is shown in table 1. Three diŒerent types of trucks, having a capacity of 2, 4 and 10 tons respectively can be used. However, up to four trucks of each type can only be operated by each carrier. Table 2 shows the passenger car equivalence rates, operating costs and ® xed costs for each type of pickup /delivery truck. These costs are based on results from recent studies of truck operations in Japan. The number of customers for each carrier was randomly generated between 5 and 24 (table 1 ). The actual nodes to be visited for each carrier were also determined randomly from all nodes in the network. Three types of time windows were permitted in this study, time windows with one hour, time windows at 09:00-12:00 or 13:00-17:00 hours and no time window. The type and starting time of each customers time window was based on a recent survey in Kobe and Osaka area. The average travel time on each link for the scanning interval is provided by the dynamic tra c simulation. In this study the scanning interval used was 30 min. When initially calculating the optimal routes and schedules, the average travel times on each link were assumed to be equal to the travel times using free running speeds. The dynamic tra c simulation requires information on passenger car behaviour, as well as optimal routes and schedules of pickup/delivery trucks, produced by the VRPTW model. This includes the departure time and order of visiting customers. Passenger cars in this study include actual passenger cars and trucks other than those that are considered in the optimal routing and scheduling model. Passenger car origin-destination (OD ) tables for every hour were estimated using tra c generation

Table 1.

Location of depot and number of customers for each freight carrier.

Freight carrier

Depot node number

Number of customers

19 13 3 24 1 2 15 6 18 17

8 22 11 17 18 15 5 19 10 20

A B C D E F G H I J

Table 2. Capacity of truck (ton) 2 4 10

Characteristics of pick up/delivery trucks. Passenger car equivalence (pcu/vehicle ) 1 1.5 2

Operating cost (yen/10 min )

Fixed cost (yen/day )

140.2 175.4 232.7

10417.5 11523.1 13789.7

78


rates at each centroid and the probability of O-D choice. The number of passenger cars for each hour was generated using a temporal demand pattern based on a tra c census conducted in Kobe City. The probability of O-D choice was estimated using: pij 5

qi qj l2ij

( 9)

where pij


qi lij

probability of choosing centroid j for passenger car tra c generated at centroid i attraction rate of centroid i distance between centroid i and centroid j

The highest level of attraction rate ( qi 5 4) was set for centroid (node ) 18 in ® gure 7, since this is assumed to be a central business district. Surrounding centroids 13, 14, 15, 17, 19, 20, 23 and 24 were set at the second level ( qi 5 3) , outer centroids 8, 9, 10, 11, 12, 16, 22 and 25 at the third level ( qi 5 2) and others at the fourth level ( qi 5 1) . The recursive relationship between the dynamic tra c simulation and the optimization of VRPTW (® gure 3 ) requires a stopping criterion for obtaining a solution. The following equation was used:

i

k

Tkin 2 Tkin2 Tkin

1

2

£ 0.05

( 10)

where Tkin is the travel time on link k for time interval i at nth iteration 3.2. Results 3.2.1. Simulating current tra c conditions. First, hypothetical tra c conditions were simulated to provide a benchmark for estimating the bene® ts of introducing the advanced routing and scheduling systems. The optimization model for VRPTW was applied to the test network. The value of the objective function for the chosen solution was 1.2 ± 1.5 times higher than that of the best solution and the average load factor of trucks was ~ 20% lower than the best solution. This discrepancy was based on the survey on the improvements found by several freight companies in Kobe City. This solution is assumed to represent the current pickup/delivery truck operations before introducing advanced systems. The pickup/delivery truck tra c was estimated to account for 14% of all tra c within the network. This percentage of pickup /delivery trucks is almost identical to the actual conditions within the Kobe area. 3.2.2. EŒects of introducing advanced routing and scheduling system. The eŒects of freight carriers introducing Advanced Routing and Scheduling System (ARSS ) on road tra c was investigated. The ARSS provides optimal routes and schedules using the VRPTW model described in the previous section. Three cases were considered, ARSS penetration at 0, 50 and 100% . In the case of 50% penetration rate, freight carriers A, B, C, D and E shown in table 1 introduced ARSS. The demand for freight transport at each customer was increased to 1.5 and 2.0 times the base case. This increase by 1.5 and 2.0 times the base case is taken as parameters. For example, if the


79


freight demand was annually increased by 2% , the demand would increase to be 1.5 times the base case in 20 years and two times in 35 years. The demand for freight

Figure 8.

EŒects of penetration rate of advanced routing and scheduling system on change in CO2 emissions with increasing demand for freight transport.

Figure 9.

EŒects of penetration rate of advanced routing and scheduling system on change in normalized CO2 emissions with increasing demand for freight transport.


80


transport in the base case was hypothetically set, as the distribution of demand coincides with the actual demand distribution given by survey in Osaka-Kobe area. Figures 8 and 9 show the eŒects of the penetration rate of ARSS on CO2 emissions with the demand for freight transport. In ® gure 9 both the change of CO2 emissions and the demand for freight transport values are normalized by the value of base case. CO2 emissions were determined using an established fuel consumption relationship. Fuel consumption was estimated using the average travel speed of vehicles on each link. This estimation was carried out for passenger cars and pickup/delivery trucks respectively and then combined together. Figure 9 indicates that the normalized CO2 emissions for the 100% penetration rate was reduced by 8.3% from that for the 0% penetration rate, when the demand was doubled. However, the normalized CO2 emissions increased by 14% when the penetration rate was 50% compared with a zero penetration rate, when the demand was doubled. This is attributed to an increased use of larger trucks that produce more CO2 emissions than smaller trucks. Freight carriers generally intend to use larger trucks, which allows them to reduce the costs compared with using smaller trucks for carrying the same amount of goods. Table 3 shows the travel time for diŒerent types of trucks. Trucks with capacity of 10 tons travelled longer periods than the other smaller trucks when the penetration rate was 50% and the demand was double the base case. In table 3 the total travel times of trucks for penetration rates of 50% and 100% are lower than that for the penetration rate of 0% in three cases of normalized demand by 7.1-14.7% . Therefore, it can be noted that ARSS considerably helps alleviate tra c congestion. The reduction of travel time of both pickup/delivery trucks and passenger cars is estimated to be 1.5 ± 2.3% . 3.2.3. EŒects of cooperative freight transport system. There are various types of cooperative freight transport systems, for example, cooperation in building and operating a common depot, cooperation in carrying goods by common pickup/ Table 3.

Change in travel time of diŒerent types of trucks by advanced routing and scheduling system. Normalized demand for freight transport

Penetration rate of ARSS (% ) 0

Capacity of truck (ton)

1.0

1.5

2.0

2 4 10

1155 1249 743

958 1424 1138

640 1291 1524

3147

3520

3455

1600 608 713

930 1200 986

732 895 1584

2921

3116

3211

947 1164 774

1039 851 1129

623 976 1349

2885

3019

2948

Subtotal 2 4 10

50

Subtotal 100

2 4 10 Subtotal

(unit: min).

81



Figure 10. Change in customers to be visited by introducing a cooperative freight transport system.


82


delivery trucks and cooperative use of information systems. Here, cooperation in carrying goods is examined as shown in ® gure 10. This ® gure demonstrate s cooperation between two freight carriers D and H, with each freight carrier having numerous customers to visit. In this cooperative freight transport system, each freight carrier collects goods from customers within its neighbourhood. As a result the total travel distance and the required number of trucks will be reduced. Here, it is assumed that ARSS is fully used by all ten freight carriers in both cases with and without cooperative freight transport. Figures 11 and 12 show the eŒects of cooperative freight transport system on the CO2 emissions by freight carriers D and H who participate in the system with the increase of demand for freight transport. These ® gures demonstrate that the CO2 emissions produced by freight carriers who participated in the cooperative system can be considerably reduced. The level of CO2 emissions produced by these freight carriers involved in cooperation remains at almost at the same level as the base case when doubling the demand for freight transport, while it doubles from the base case without cooperation. The normalized CO2 emissions with cooperation were reduced by 51.8% from that without cooperation, when the demand was doubled. Figure 11 indicates that the CO2 emissions in the base case without cooperation are lower than those with cooperation. This is due to the increased travel times experience by trucks with the capacity of 10 tons as shown in table 4. Table 4 also shows that total travel times for all truck types are reduced in all of the demand cases considered. This produces bene® ts relating to better tra c ¯ ow conditions on the network. Table 5 shows the total costs of freight carriers D and H. The total costs are reduced by 2329% after implementing a cooperative freight transport system for the three demand levels. Therefore, cooperative freight transport systems are eŒective in reducing costs at various levels of demand. Figure 13 shows the CO2 emissions produced by all

Figure 11. EŒects of cooperative freight transport system on change in CO 2 emissions by freight carriers D and H with increasing demand for freight transport.

83



freight carriers within the network. This ® gure indicates that the CO2 emissions produced by all freight carriers are reduced for the normalized demand for freight transport of 1.0 and 2.0 by cooperative freight transport, but are slightly higher for the normalized demand for freight transport of 1.5. This is again due to the increase of travel time of trucks with the capacity of 10 tons as shown in table 6. In table 6 the total travel time of all types of trucks for all freight carriers is reduced by 4-10% after

Figure 12. EŒects of cooperative freight transport system on change in normalized CO2 emissions by freight carriers D and H with increasing demand for freight transport.

Table 4.

Change in travel time of diŒerent types of trucks for freight carriers D and H by cooperative freight transport system. Normalized demand for freight transport

Without cooperation (min )

With cooperation (min )

Capacity of truck (ton )

1.0

1.5

2.0

2 4 10

105 368 236

239 0 572

160 170 538

Subtotal

709

811

868

2 4 10

94 160 364

75 177 371

150 152 309

Subtotal Change by cooperation (% )

618 Ð

12.8

623 Ð

23.2

611 Ð

29.6

84 Table 5.

E. Taniguchi and R. E. C. M. van der Heijden Change in total cost of freight carriers D and H by cooperative freight transport system.

Normalized demand for freight transport Cost without cooperation (yen ) Cost with cooperation (yen )

1

1.5

2

154337 118218

190135 139257

225996 159853


Change by cooperation (% )

23.4 Ð

Ð

26.8 Ð

29.3

Figure 13. EŒects of cooperative freight transport system on change in CO2 emissions by all freight carriers with increasing demand for freight transport.

Table 6.

Change in travel time of all freight carriers by cooperative freight transport system. Normalized demand for freight transport

Without cooperation

With cooperation

Change by cooperation (% )

Capacity of trucks (ton)

1.0

1.5

2.0

2 4 10

947 1164 774

1039 851 1129

623 976 1349

Subtotal

2885

3019

2948

2 4 10

1031 798 751

746 834 1236

797 806 1226

Subtotal

2580

2816

2829

Ð

10.6 Ð

6.7 Ð

4.0

85


implementing the cooperative system. It is remarkable that the cooperation of only two companies out of ten produces such large bene® ts. 3.2.4. EŒects of controlling load factor. An initiative of controlling load factors of pickup /delivery trucks is examined in this section. The load factor of a truck is de® ned as follows: Wl,i Fl 5

l

i

( 11)

Wc,l . nl

l


where Wl,i Wc,l nl

load of vehicle l with a load at i -th trip (kg ) capacity of vehicle l (kg ) number of trips by vehicle l with a load

This load factor incorporates only vehicles travelling with a load. The load factor is de® ned in terms of weight in this paper. Since the term weight is used in most freight transport statistics, and the weight indicates the real weight for heavier goods, and the converted weight from volume for lighter goods. First the load factor of the base case was calculated, in which ARSS is fully used by ten freight carriers, but no regulation for load factors was applied. The average load factor de® ned by equation (11 ) that excludes vacant truck was 37.6% in base case, ranging from 33.9% for the worst freight carrier to 45.1% for the best freight carrier. Then, three cases of regulating the load factor were investigated: 1. 2. 3.

Case 1: average load factor must be > 35% . Case 2: average load factor must be > 37.5% . Case 3: average load factor must be > 40% .

Results showed that the average load factor increased to 43.9% for case 1, 44.8% for case 2, and 46.1% for case 3 (table 7 ). The average load factor was eventually improved from base case by 6.3, 7.2 and 8.5 points for cases 1-3 respectively. Table 7 Table 7.

Load factor of pickup/delivery trucks. Normalized demand for freight transport

Regulation

Type of load factor

1

1.5

2

No regulation

average (including vacant truck ) average (excluding vacant truck ) maximum

22.6 30.5 46.1

24.2 34.3 46.9

25.9 47.2 51.1

> 35%


28.2 38.7 53.4

27.2 38.4 55.4

30.5 43.9 60.4

> 40%


n.a. n.a. n.a.

31.0 42.6 58.3

30.1 46.1 64.2

(unit: % )


86


shows the maximum load factor that is de® ned as the average of the maximum value of the load factor of a vehicle in the travel chain from and to the depot. The maximum load factor also improved with the increase in the average load factor. Figure 14 shows the change in CO2 emissions and costs with the average load factor of pickup/delivery trucks when load factors were controlled. This ® gure indicates that the minimum CO2 emissions and costs for the four cases when the average load factor was 43.9% . It means that there is an optimal load factor for minimising both CO2 emissions and cost. Therefore, city planners should be careful when choosing an appropriate level for controlling the load factor in urban freight transport. Figures 15 and 16 show the eŒects of controlling load factors on CO2 emissions produced by pickup/delivery trucks with increasing demand for freight transport. These ® gures highlight a clear reduction of CO2 emissions by restricting the average load factor to be above a certain level. CO2 emissions increase by up to 1.7 times when the demand is double the base case without regulation, but rise only 1.4 times when the average load factor is > 35% . The reduction rate of normalized CO2 emissions is 18.2% , when the demand is doubled. Therefore, controlling the average load factor of pickup/delivery trucks is an eŒective measure to depress the increasing CO2 emissions associated with the increase of demand for freight transport. The total costs for freight carriers change slightly and are within the range of Ð 3.0 to + 3.2% (table 8 ). 4. Conclusions This paper presented a methodology for evaluating several city logistics initiatives using dynamic tra c simulation with optimal truck routing and

Figure 14. Change in CO 2 emissions and cost with average load factor of pickup/delivery trucks.

87


scheduling. This methodology was applied to a test road network and three city logistics initiatives were evaluated: (1 ) Introducing advanced routing and scheduling system, (2 ) Implementing cooperative freight transport system, and (3 ) Controlling the load factors of pickup/delivery trucks in urban areas. The main criterion for the evaluation used in this paper was CO2 emissions produced by pickup/delivery trucks when the demand for freight transport at customers was increased. The following ® ndings were derived from the evaluation of three city logistics initiatives. 1. Introducing advanced routing and scheduling systems helps reduce CO2 emissions when the demand for freight transport increased. The normalized


Table 8.

Change in cost by controlling load factor. Normalized demand for freight transport 1.0

No regulation > 35% > 40%

1.5

Cost (yen )

Change from base case (% )

Cost (yen )

511272 527556 n.a.

3.2

612970 611485 594520

2.0

Change from base case (% )

Ð

Ð

0.2 3.0

Cost (yen ) 629608 621628 640001

Change from base case (% ) Ð

1.3 1.7

Figure 15. EŒects of controlling load factor on change in CO2 emissions by trucks with increasing demand for freight transport.


88


Figure 16. EŒects of controlling load factor on normalized change in CO 2 emissions by trucks with increasing demand for freight transport.

CO2 emissions reduced by 8.3% when the penetration rate rose to 100% from 0% , when the demand level was doubled. 2. Cooperative freight transport systems can considerably reduce CO2 emissions by reducing the distance travelled by trucks. CO2 emissions produced by freight carriers with cooperation remained at almost same level as base case when the demand for freight transport was doubled, while they increased to almost double the base case without cooperation. The normalized CO2 emissions with cooperation were reduced by 51.8% from that without cooperation, when the demand was doubled. The cooperative freight transport system was eŒective not only in reducing the total costs of freight carriers but also in alleviating the environmental impacts of urban freight transport. 3. Controlling load factors of pickup/delivery trucks also produces bene® ts by reducing the total costs and CO2 emissions. Results showed that there is an optimal load factor for minimizing the total cost and CO2 emissions. This measure is expected to help depress the rate of CO2 emissions associated with the increase of demand for freight transport. The normalized CO2 emissions with the regulation of load factor were reduced by 18.2% from that without the regulation, when the demand was doubled. Therefore, all three city logistics initiatives were found to be eŒective for reducing the increase in total costs of freight carriers and the CO2 emissions produced by pickup /delivery trucks for increasing levels of demand for freight transport. The methodology presented in this paper allows city planners to quantitativel y evaluate


89

the environmental impacts of city logistics initiatives with increasing levels of urban freight transport.


References BRAMEL , J. and SIMCHI-LEVI, D., 1996, Probabilistic analysis and practical algorithms for the vehicle routing problem with time windows. Operations Research, 44, 501 ± 509. DUIN, J. H. R. VAN, 1997, Evaluation and evolution of the city distribution concept. In: 3rd International Conference on Urban Transport and the Environment for the 21st Century, Terni, Italy, pp. 327 ± 337. DUIN?, J. H. R. VAN, 1998, Simulation of underground freight transport systems. In: 4th International Conference on Urban Transport and the Environment for the 21st Century, Lisbon, Portugal, pp. 149 ± 158. EUROPEAN COMMISSION, DIRECTORATE GENERAL FOR TRANSPORT 1997, Urban Goods Transport. Final Report of COST 321. FUJII, S., IIDA , Y. AND UCHIDA, T., 1994, Dynamic simulation to evaluate vehicle navigation. In: Vehicle Navigation & Information Systems Conference Proceedings, pp. 239 ± 244. GENDREAU, M., LAPORTE, G. AND POTVIN, J. ± Y., 1997, Vehicle routing: modern heuristics. In: E. Aarts and J. K. Lenstra (eds), Local Search in Combinatorial Optimization (Chichester: John Wiley ), pp. 311 ± 336. JANSSEN, B. J. P. AND OLDENBURGER, A. H., 1991, Product channel logistics and city distribution centers; the case of the Netherlands. In: OECD Seminar on Future Road Transport Systems and Infrastructures in Urban Areas, Chiba, Japan, pp. 289 ± 302. KOHLER, U., 1997, An innovating concept for city ± logistics. In: 4th World Congress on Intelligent Transport Systems, Berlin, Germany, CD ± ROM. KOKUBUGATA , H., ITOYAMA , H. AND KAWASHIMA, H., 1997, Vehicle routing methods for city logistics operations. In: M. Papageorgiou and A. Pouliezos (eds), IFAC/IFIP/IFORS Symposium on Transportation Systems, Chania, Greece, pp. 755 ± 760. KOSHI, M., YAMADA , H. AND TANIGUCHI E., 1992, New urban freight transport systems. In: Selected Proceedings of 6th World Conference on Transport Research, Lyon, France, pp. 2117 ± 2128. KOSKOSIDIS, Y. A., POWELL , W. B. AND SOLOMAN M. M., 1992, An optimization ± based heuristic for vehicle routing and scheduling with soft time window constraints. Transportation Science, 26, 69 ± 85. OISHI, R., 1996, A study on planning methodology for new urban freight transport systems. PhD dissertation, Kyoto University. [in Japanese] POTVIN, J. ± Y., KERVAHUT, T., GARCIA , B. ± L. AND ROUSSEAU, J. ± M., 1996, The vehicle routing problem with time windows; Part I: Tabu search. INFORMS Journal on Computing, 8, 158 ± 164. RUSKE, W., 1994, City logistics solutions for urban commercial transport by cooperative operation management. In: OECD Seminar on Advanced Road Transport Technologies, Omiya, Japan. RUSSELL, R. A., 1995, Hybrid heuristics for the vehicle routing problem with time windows. Transportation Science, 29, 156 ± 166. SOLOMON , M. M., 1987, Algorithms for the vehicle routing and scheduling problems with time window constraints. Operations Research, 35, 254 ± 265. TANIGUCHI, E., YAMADA , T. AND YYANAGISAWA , T., 1995, Issues and views on cooperative freight transport systems. In: 7th World Conference on Transport Research, Sydney, Australia. TANIGUCHI, E., NORITAKE, M., YAMADA , T. AND IZUMATANI, T., 1997, Optimizing the size and location of logistics terminals. In: IFAC/IFIP/IFORS Symposium on Transportation Systems, Chania, Greece, pp. 741 ± 746. TANIGUCHI, E., YAMADA , T., TAMAISHI , M and NORITAKE , M., 1998, EŒects of designated time on pickup/delivery truck routing and scheduling. In: 4th International Conference on Urban transport and the Environment for the 21st Century, Lisbon, Portugal, pp. 127 ± 136.

90



THANGIAH, S. R., NYGARD , K. E. and JUELL, P. L., 1991, GIDEON: a genetic algorithm system for vehicle routing with time windows. In: 7th IEEE International Conference on Arti® cial Intelligence Applications (Los Alamitos: IEEE Computer Society Press ), pp. 322 ± 328. VISSER, J. G. S. N., 1997, Underground networks for freight transport: a dedicated infrastructure for intermodal short ± distance freight transport. In: Proceedings of 3rd TRAIL Year Congress on Transport, Infrastructure and Logistics, Delft, The Netherlands.