Conference PaperPDF Available

LS/ATN: Reporting on a Successful Agent-Based Solution for Transport Logistics Optimization

Abstract and Figures

A considerable volume of research exists concerning the domain of automatic planning and scheduling, hut many real-world scheduling problems, and especially that of transportation logistics, remain difficult to solve. In particular, this domain demands schedule-solving for every vehicle in a transportation fleet where pick-up and delivery of customer orders is distributed across multiple geographic locations, while satisfying time-window constraints on pickup and delivery per location. This paper presents a successful commercial-grade solution to this problem called living systems adaptive transportation networks (LS/ATN), which has been proven through real-world deployment to reduce transportation costs through the optimization of route solving for both small and large fleets. LS/ATN is a novel agent-based resource management and decision system designed to address this highly dynamic and complex domain in commercial settings. We show how LS/ATN employs agent cooperation algorithms to derive truck schedules that optimize the use of available resources leading to significant cost savings. The solution is designed to support, rather than replace, the day-to-day activities of human dispatchers
Content may be subject to copyright.
LS/ATN: Reporting on a Successful Agent-Based Solution for Transport
Logistics Optimization
Nicoleta Neagu and Klaus Dorer and Dominic Greenwood and Monique Calisti
Whitestein Technologies AG.
Pestalozzistrasse 24
8032 Zurich, Switzerland
nne, kdo, dgr,
A considerable volume of research exists concerning the
domain of automatic planning and scheduling, but many
real-world scheduling problems, and especially that of
transportation logistics, remain difficult to solve. In par-
ticular, this domain demands schedule-solving for every
vehicle in a transportation fleet where pick-up and deliv-
ery of customer orders is distributed across multiple geo-
graphic locations, while satisfying time-window constraints
on pickup and delivery per location.
This paper presents a successful commercial-grade so-
lution to this problem called Living Systems Adaptive
Transportation Networks (LS/ATN), which has been proven
through real-world deployment to reduce transportation
costs through the optimization of route solving for both
small and large fleets. LS/ATN is a novel agent-based re-
source management and decision system designed to ad-
dress this highly dynamic and complex domain in commer-
cial settings. We show how LS/ATN employs agent coop-
eration algorithms to derive truck schedules that optimize
the use of available resources leading to significant cost
savings. The solution is designed to support, rather than
replace, the day-to-day activities of human dispatchers.
1 Introduction
The global transportation logistics sector is growing and
consolidating day by day. European operators are joining
forces with one another and with counterparts around the
world. This growth only leads to exacerbating what is al-
ready a complex problem to solve: how to efficiently in-
tegrate distributed dispatching centers to bring about the
means to strategically plan network operations using route
Any solution must account for the considerable volume
of communication and remote coordination necessary to
detect any opportunities of combining transportation be-
tween regional dispatching centers. More precisely, delays
and coordination costs can significantly impact the chance
to optimize resource consumption (i.e., deployed vehicles,
number of dispatching centers) and finally fulfil all orders
as expected. For instance, the average utilization of ve-
hicles in a truck charter business is fairly low (as low as
55% [1]), partially because order consolidation between
regions are missed. Furthermore, while a number of exist-
ing computational solutions allow the automatic creation of
the dispatching plans, they typically do not, or only mar-
ginally, support the case of plan deviations resulting from
unexpected, or previously unknown events. Naturally this
problem grows by factors as companies expand their op-
erations and integrate with others while constantly dealing
with shrinking margins and substantial cost pressures. In
response, this paper reports on a major step toward solv-
ing these problems through a new software agent-based ap-
plication, the Living Systems Adaptive Transportation Net-
works (LS/ATN)
solution that has now been successfully
deployed by two major transportation logistics companies,
with third now underway. The latter of these installations
will mean that the software will be in use across 17 coun-
tries, dealing with more than 15,000 trucks and 40,000 or-
ders each day
. This will probably be the largest commer-
cial deployment of agent technology in the world.
The LS/ATN solution is the result of several years of
investment into learning how to harness the flexible and
adaptive nature of software agent technology in applica-
tion to the core scheduling optimization problems in the
transportation logistics domain. As an intelligent trans-
port optimization system, LS/ATN is able to dynamically
adapt transport plans and scheduling according to possi-
For further information please go to
es/solutions/logistics/ls atn.html
Approximate numbers.
ble deviations and unforeseen events. This is achieved by
solving the classic operations research dynamic multiple
Pickup and Delivery Problem With Time Windows prob-
lem (mPDPTW) [4, 7, 8, 9], which is described briefly in
Section 2. Several agent-based approaches have been pro-
posed to deal with this kind of dynamic optimization prob-
lem, e.g. [3, 6].
The most recent success story of LS/ATN is an instal-
lation with ABX Logistics, who with a staff of around
10,000 in 30 countries, represent one of the world’s leading
transport and logistics service providers. Each day ABX
dispatchers manage several hundreds of orders and sev-
eral hundreds of trucks within the German full truck load
and part truck load sector. A careful analysis of ABX’s
flow of goods detected significant potential for optimiza-
tion thatwould lead toreduced driven-emptykilometers and
thereby overall costs. Other stated requirements for the so-
lution were (1) full visibility for all users, (2) exploitation of
the potential for global optimization, (3) simplification and
standardization of the dispatching and settlement processes,
(4) immediate reaction to events such as order changes, or-
der cancelations or other plan deviations and (5) generation
of up-to-date business figures (KPIs) to ensure the highest
quality of service.
After convincing ABX that software agents were a
technically sound approach to achieving good results, the
LS/ATN solution was installed with substantial impact. The
LS/ATN system is now running successfully for several
months. LS/ATN software agents support the dispatchers
by performing optimizationtasks followinga bottom-up ap-
proach in first searching for local optima and then, in a sec-
ond step, enlarging the search space to seek approaching
more the global optimum. Results reported to date indicate
that LS/ATN has delivered an estimated 30% improvement
in process efficiency (see 1) for ABX resulting in a signifi-
cant reduction of transportation costs.
The paper is organized as follows: Section 2 covers the
mPDPTW problem domain and related work, Section 3 de-
scribes the LS/ATN system architecture and optimization
approaches, Section 4 reports and discusses the empirical
results obtained by solving a real world optimization prob-
lem and finally, in Section 5 we conclude the paper.
2 The mPDPTW Problem Domain
Our research concerns the multiple Pick up and Delivery
Problem With Time Windows (mPDPTW) and is motivated
by its application to the transport logistics domain. The
mPDPTW problem consists of computing the optimal set of
routes for a fleet of vehicles in order to satisfy a collection
of transportation orders while satisfying the time windows
at client locations. Each order includes a pickup and a deliv-
ery location, andtime windowsassociated with each service
location within which the order has to be picked up or deliv-
ered. Vehicles are located in various initial positions (there
is no central depot) and must be dispatched and routed so
that each request is picked up at its origin before being de-
livered to its destination. The goal of the mPDPTW is to
provide feasible schedules, which satisfy the time window
constraints, for each vehicle in order to deliver to a set of
customers with known demands on minimum-cost vehicle
While a significant amount of research exists in the do-
main of planning and scheduling, the problems of vehi-
cle routing and order scheduling are far from adequately
solved in practice. This is due to the fact that in research
studies many parameters or customer requests are ignored.
Moreover, although there exist efficient algorithms for solv-
ing static scheduling problems where all the data about the
client orders is knownin advance, in practice one has to deal
with dynamic scheduling, where client orders or changes in
already requested orders might arrive in the system at any
time. Distributed aspects of planning and scheduling prob-
lems are hard to define in a generic manner since they often
depend on the application domain.
Today, several industries including transportation logis-
tics are faced with constantly spreading world-wide trading
and goods flow. This global context requires distribution
and high flexibility in the transportation scheduling system,
which can be achieved through the application of multi-
agent systems, as demonstrated by other approaches such
as [3, 6]. The specific needs of the transportation domain
computation can classified as follows:
Distribution: When system capabilities are set apart
into independent units/agents they may be intrinsically
distributed over a large network of computers.
Task decomposition: Transportation applications are
especially suitable for the application of techniques
such as task decomposition, where the schedule for
each vehicle is computed by a single agent.
Decentralized scheduling: Computation and system
control are distributed among the agents. Each agent
can be designed to act independently by computing a
part of the schedule without needing knowledge,or rea-
soning, about the global process of the whole system.
Cooperation among the agents: In order to achieve a
better global solution, the agents must cooperate by ex-
changing client orders between one other and adjusting
their schedules accordingly with the goal of minimiz-
ing the overall cost.
Based on our practical experience with transportation
scheduling in medium and large-size logistics companies,
we can testify to the suitability of these techniques to real-
world problems.
Apart from the advantages given by the use of a multi-
agent architecture for solving the mPDPTW problem, in
real-life applications we need to handle also the dynamic
aspects of this problem. For solving a mPDPTW prob-
lem dynamically, we distribute it among multiple inter-
acting agents in order to (1) achieve scalability of perfor-
mance with growing sizes of problem instances; (2) di-
rectly reflect the distributed nature of transportation net-
works/organizations and decision making centers; (3) fa-
cilitate the handling of local deviations without the need
to propagate local changes and recompute the whole solu-
tion, and (4) increase robustness (avoiding single point-of-
Several agent-based systems have been proposed in or-
der to distribute the computation. The most radical one is
to represent each vehicle by an agent [3, 6]. Solution gen-
eration is done by sequential insertion of orders handled by
a contract-net interaction protocol (see, [3]). Optimization
can be achievedby triggering order transfers between trucks
whenever this improves the objective function. In such a
fully distributed architecture, vehicle agents with a changed
route start an optimization process in parallel to the other
active vehicle agents. The main advantage of such an ap-
proach is in its fine granularity and high scalability, while
its main disadvantage stems from a considerable overhead
in computation time and resource usage. The overhead in
computation time is mostly due to more expensive agent
communications when compared to a fully centralized so-
lution, while the overhead in resource usage depends on the
memory and processing footprint of an agent.
3 The LS/ATN Approach
An agent-oriented approach to transport optimization
system is proposed. First we present how LS/ATN is in-
tegrated into the main dispatching process followed by the
agent-based architecture and optimization of the system.
3.1 LS/ATN Process Integration
In Figure 1, we show the integration of LS/ATN into the
main dispatching process and how it interacts with the trans-
portation environment.
Client orders are received by the LS/ATN system
through communication with the Transportation Manage-
ment System (TMS) Order Entry functionality. The client
data is processed by the Dispatching Support module of the
LS/ATN system. Routes are computed based on agent co-
operation as described in Figure 3, and improved by agent-
based optimization algorithms, see Section 3.3. The plan
obtained from this collaborative processing is then reported
Figure 1. Main process of LS/ATN.
to the Dispatcher for execution, and optionally for manual
dispatching. The final routes accepted for execution and
potentially adapted with manual dispatching are ordered for
execution to a carrier and reported to the Tracking module
of the LS/ATN system. The tracking module supports track-
ing and event handling for orders and trucks during the ex-
ecution phase of the transport. Once this is finished it sends
the routes back to the Dispatcher for post execution admin-
istrative operations. The final decisions regarding costs are
reported to the Accounting module of the TMS system.
3.2 LS/ATN Agent System Architecture
In Figure2, we present the main architecture of our trans-
port logistics system. The agents used within the system
can roughly be divided into two groupes: communication
agents allowing to interface to external modules like the
clients, geo-coding information systems
, transport man-
agement systems, telematics systems and other external sys-
tems and we have optimization agents described in the next
section. The agent-based architecture is very flexible and
thus, highly extendable. It allows easily for the introduc-
tion of new agents occurring as a result of customer needs
or system evolution.
3.3 LS/ATN Agent-Based Optimization
The agent design chosen for optimization directly re-
flects the way logistics companies actively manage the com-
plexity of this domain. The global business is divided into
regional businesses which are usually dispatched in distrib-
uted dispatching centers. This distribution is represented by
We are using a PTV servers for geographical services, see:
Figure 2. LS/ATN system architecture.
agents communicating to each other. In the following we
describe how automatic route planning is handled by our
system based on agent collaboration, step 2 in Figure 1.
3.3.1 Agent Cooperation for Routes Solving
The transportation business is usually divided into re-
gions/clusters. Transportation requests arriving at a cluster
are first tentatively allocated and possibly optimized within
that cluster. If the order
s pickup or delivery location is in a
differentcluster, the other cluster is alsoinformed and asked
to handle the request if it can do so in a cheaper way. In
our agent-based framework, distinct software agents repre-
sent different regional clusters. All vehicles starting in the
region of an agent are managed by a local AgentCluster-
Manager. Incoming transport requests are distributed by a
centralized AgentDistributor according to their pickup lo-
cation (see Figure3). Clusters may also be configured to be
managed dynamically containing all trucks that have routes
passing the cluster allowing the clusters to overlap.
The main advantage of this latter design stems from its
direct mapping to todays transport business organization
and its good scalability. The computational overhead in-
curred by such a multi-agent based solution is also much
lower than that occurring in a fully distributed solution.
Nevertheless, besides degradation of the solution quality
when compared to a fully centralized approach, the main
disadvantage (also with respect to the fully distributed op-
tion) is that optimization within a cluster and among clusters
has to be handled slightly differently.
3.3.2 Optimization Algorithms
The optimization process incrementally reflects the dynam-
ics of the underlying mPDPTW settings. Whenever a new
Figure 3. A summarized view of a cluster-
based agent design.
transport request is made available to the system, the cur-
rent delivery plan is updated. This is done in a two-phase
approach: First, a new valid solution is generated including
the new transportation request. Then, the obtained solution
is improved by negotiation between the agents to transfers
orders in case this reduces the overall costs.
The algorithm used for assigning an order to a truck is a
sequential insertion of orders [5]. All available trucks under
control of the AgentClusterManager are checked to see if
they are able totransport the order and what additional costs
are incurred. The order is finally assigned to the truck with
the least additional costs.
Sequential insertion with requests for quotes to all trucks
potentially produces suboptimal solutions. See for instance
the example given in Figure 4. Order 1 is the first to arrive
in the system and is assigned to truck 1’s route. Order 2
is also optimally assigned to truck 1’s route since that pro-
duces least additional costs (and kilometres). When order
3 arrives truck 1 is fully loaded, therefore a new truck 2 is
used for order 3 and later for order 4.
In order to improve the solution a further optimization
step is performed by cyclic transfers between trucks. A
cyclic transfer is an exchange of orders between routes. Fig-
ure 5 shows how the suboptimal example in Figure 4 is im-
proved by a transfer of order 2 from route 1 to 2 while order
4 is transferred from route 2 to 1.
The optimization procedure must determine which trans-
fers should be triggered. The AgentClusterManager there-
fore starts a negotiation process withthe truck that was most
recently changed. That truck is initiating transfer requests
to all other trucks under control of AgentClusterManager.
From all the requests the most significant cost-saving trans-
fer is performed. This changes the routes of both trucks
Delivery all orders
Pickup order 1
Pickup order 2
Pickup order 3
Pickup order 4
Truck 1 start
Truck 2 start
Figure 4. Example of a suboptimal solution
given by the insertion algorithm for 4 orders.
Delivery all orders
Pickup order 1
Pickup order 2
Pickup order 3
Pickup order 4
Truck 1 start
Truck 2 start
Figure 5. Optimal solution, for the example
given in Figure 4, after an order transfer.
involved. This hill climbing process is then continued with
all changed routes until no more cost-saving exchanges can
be achieved.
4 Discussion and Results
There are various aspects which are requested for op-
timization in a transport logistic system and considered in
LS/ATN: transportation cost, solution quality, process cost,
more flexibility for dispatchers and cost transparency. We
discuss how these parameters are treated by our system for a
major European logistics company. We also report how the
results are improved by automatic optimization provided by
our system in comparison with manual dispatching used by
the transport logistic companies today. Because of space
limitations we present our results only relative to the first
two parameters aforementioned. The given dataset contains
roughly 3500 real-business transportation requests.
Evaluation Parameter 1 Saving (agent-based)
overall cost 11.7%
driven kilometers 4.2%
deployed vehicles 25.5%
Table 1. Savings achieved by the agent-based
approach when compared to manual dis-
The major goal of logistics companies is reduction in
transportation costs. This can be achieved by higher uti-
lization of transportation capacity. This results in a reduced
number of driven kilometers andless trucks needed. Table 1
summarizes the comparison of these major results gained by
comparing the solution of manual dispatching with process-
ing the same orders on ATN. A total of 11.7% cost savings
was achieved, where 4.2% of the cost savings stem from
an equal reduction in driven kilometers. An additional im-
portant achievement is that the number of vehicles used is
25.5% lower compared to the manual solution. The cost
savings would even be higher if fixed costs for the vehicles
were feasible, which is not the case in the charter business,
but possibly in other transportation settings.
Our agent-based optimization system guarantees a
higher service level in terms of results quality. The high so-
lution quality corresponds to a reduced number of violated
constraints. Our system allows for fine tuning of the desired
level of service quality. Figure 6 presentsvarious results ob-
tained by our system relative to the solution proposed by
manual dispatching. The first proposed solution (ATN1)
provides a reduction of 8.33 % in driven kilometers at the
same service level with no more than 25% of violated con-
straints with more than 6h pickup or delivery delay. More-
over, this solution provides also a reduction of 8% in terms
of kilometers driven with empty trucks. The second solu-
tion (ATN2) proposes a reduction in driven kilometers of
0.78% relatively to the manual dispatching solutions, while
providing a significant higher service level: only 2.5% of
violated constraints with more than 6h delay. The third so-
lution (ATN3) proposes an increase of only 1.66% in terms
of driven kilometers, while meeting all the constraints.
Through the use of the automatic optimization a lower
process cost is being achieved. This is due to automatic han-
dling of plan deviations and evaluation of solution options
in real-time. Moreover, through automation the communi-
cation costs in terms of dispatcher
s time and material is re-
duced. Better customer support can be guaranteed through
fast, comprehensive and up-to-date information about order
Automation also allowsprocessing of a higher number of
orders than with manual dispatching. This is an important
issue as the size of data to be managed is increasingly grow-
Figure 6. A summarized view of a cluster-based agent design.
ing. Other advantages of using LS/ATN are: cost trans-
parency and seamless integration with TMS and telematics.
5 Conclusions
Interest in distributed agent-based decision systems for
dynamic, unpredictable and distributed application domains
such as transport logistics is increasing significantly as hu-
man dispatchers begin to be overwhelmed by the amount
of data needed to be handled. We present LS/ATN as a
solution to this problem that has been proven in use by a
growing number of major transport companies. The de-
sign of LS/ATN as a software agent application is motivated
largely by the high responsiveness of autonomous agents as
entities that can react locally to changes in complex envi-
ronments. Moreover, the agent-oriented paradigm for soft-
ware engineering provides a strong basis for the construc-
tion of large, complex systems, in which components can
be naturally distributed across a network of heterogenous
computers, without demanding a complete analysis of their
Our solution, LS/ATN, is a commercial system for com-
puting the truck schedules in transportation logistic applica-
tions. We are using multi-agent technologies as the platform
underlying our system and we are experimenting with vari-
ous techniques for schedule solving. Our production system
is now successfully used by ABX logistics and several other
medium to large-sized transport logistics companies. While
ABX has achieved a reduction of 11.7% in costs relative to
the manual dispatchers solution, we typically guarantee a
reduction of at least 4% to 6%. This improvement is sig-
nificant for transport companies that have huge number of
orders to manage and significant costs, but small profit mar-
[1] K. Dorer and C. Calisti. Agent-based Dynamic Transport Op-
timization. In Whitestein Technologies Technical Report, WT-
2004-05, 2004.
[2] K. Dorer and C. Calisti. An Adaptive Solution to Dynamic
Transport Optimization. In Proceedings of the AAMAS05 in-
dustry track, Utrecht, The Netherlands, 2005.
[3] K. Fischer. Cooperative Transportation Scheduling: an ap-
plication Domain for DAI. In Journal of Applied Artificial
Intelligence, vol. 10, pp. 1–34, 1995.
[4] M. Gendreau and J. Potvin. Dynamic vehicle routing and dis-
patching. T. Crainic and G. Laporte, editors, Fleet Manage-
ment and Logistics, Kluwer, 33(4):115126, 1998.
[5] J.-J. Jaw, A. R. Odoni, H. N. Psaraftis, and N. Wilson. A
heuristic algorithm for the multivehicle advance request dial-
a-ride problem with time windows. Transportation Research,
vol. 20 B (3), pp. 243–257, 1986.
[6] R. Kohout and K. Erol. In-time agent-based vehicle routing
with a stochastic improvement heuristic. In Proceedings of
the 6th National Conference on Artificial Intelligence (AAAI-
99); Proceedings of the 11th Conference on Innovative Ap-
plications of Artificial Intelligence, pages, AAAI/MIT Press.,
page 864869, Menlo Park, Cal, 2004.
[7] S. Mitrovic-Minic. Pickup and delivery problem with time
windows: A survey. In Technical Report TR 1998-12, School
of Computing Science, Simon Fraser University, Burnaby,
BC, Canada,, 1998.
[8] M. W. P. Savelsbergh and M. Sol. The general pickup and de-
livery problem. Transportation Science, 29(1):17–29, 1995.
[9] C. T.G. Long haul freight transportation. R.W. Hall, editor,
Handbook of Transportation Science, 2002.
... Cuenca newtork links used to generate the metamodel ED as ranked by link-betweeness. optimal recovery and adaptation decision-making, such as updates to traffic models using a Bayesian approaches Castillo et al. (2008), or agent-based analyses, such as LS/ ATN Dorer and Calisti (2005); Neagu et al. (2006). The present implementation is representative of the enabling potential of metamodeling in one of the most relevant topics of the present, recovery, and adaptation in the context of engineering systems. ...
Full-text available
Different emerging threats highlighted the relevance of recovery and adaptation modelling in the functioning of societal systems. However, as modelling of systems becomes more complex, its effort increases challenging the practicality of the engineering analyses required for efficient recovery and adaptation. In the present work, metamodels are researched as a tool to enable these analyses in traffic networks. One of the main advantages of metamodeling is their synergy with the short decision times required in recovery and adaptation. A sequential global metamodeling technique is proposed and applied to three macroscopic day-to-day user-equilibrium models. Two reference contexts of application are researched: optimal recovery to a perturbation (with response times reduced by 98% with loss of accuracy lower than 1%) and adaptation under uncertainty with perturbation-dependent optimality. Results show that metamodeling-based metaheuristics enable fast resource-intensive engineering analyses of traffic recovery and adaptation, which may change the paradigm of decision-making in this field.
... Mobile agent oriented technology is quite popularly used in e-commerce [35], search engines [36,46], webcrawlers [33], scheduling problems [22,30], parallel computing [24], health monitoring [39], network administration [34] and many other areas [41]. The mobility nature of the agent makes it more valuable while moving from one platform to another to perform different tasks on behalf of its owner. ...
Mobile agent technology is an active research topic and has found its uses in various diverse areas ranging from simple personal assistance to complex distributed big data systems. Its usage permits offline and autonomous execution as compared to classical distributed systems. The free roaming nature of agents makes it prone to several security threats during its transit state, with an added overhead in its interoperability among different types of platforms. To address these problems, both software and hardware based approaches have been proposed to ensure protection at various transit points. However, these approaches do not ensure interoperability and protection to agents during transit over a channel, simultaneously. In this regard, an agent requires a trustworthy, interoperable, and adaptive protocol for secure migration. In this paper, to answer these research issues, we first analyse security flaws in existing agent protection frameworks. Second, we implemented a novel migration architecture which is: 1) fully inter-operable compliance to the Foundation for Intelligent Physical Agents (FIPA) and 2) trustworthy based on Computing Trusted Platform Module (TPM). The proposed approach is validated by testing on software TPM of IBM, JSR321, and jTPMTools as TPM and Trusted Computing Software Stack (TSS) interfaces, JADE-agent framework and 7Mobility Service (JIPMS). Validation is also performed on systems bearing physical TPM-chips. Moreover, some packages of JIPMS are also modified by embedding our proposed approach into their functions. Our performance results show that our approach merely adds an execution overhead during the binding and unbinding phases
... Cooperation of humans and intelligent agents for auctioning in transportation logistics is implemented in platform developed by CWI (National Research Center for Mathematics and Computer Science, Amsterdam, The Netherland) and VOS logistics companies(Nijmegen, The Netherlands) (Robu, et al., 2011). The LS/ATN system (Living Systems /Adaptive Transportation Networks) (Neagu, et al., 2006) uses agent techniques (mostly constraint-reasoning type techniques) for dynamic transport optimization. The Magenta system (Skobelev, et al., 2007) is another such system, which explores the use of swarm-based optimization techniques (Dorigo, et al., 2008). ...
Purpose In difficult geographical zones (mountain, intra-cities areas, etc.), many shippers, from small and medium enterprises to individuals, may demand delivery of different food products (fresh, refrigerated, frozen, etc.) in small quantities. On the other side, carrier companies wish to use their vehicles optimally. Taking into account the perishability constraints (short-shelflife, temperature limits, etc.) of the transported food products and environmental constraints (pollution, carbon impact) while consolidating multiple kinds of food products to use vehicles optimally is not achieved by current transportation planning solutions. The purpose of this paper is to present an interoperable solution of a marketplace, formed by shippers and carriers, dedicated to the schedule of food transport orders. Design/methodology/approach This transportation planning system named Interoperable-Pathfinder, Order, Vehicle, Environment and Supervisor (I-POVES) is an interoperable multi-agent system, based on the SCEP (supervisor, customer, environment and producer) model (Archimede and Coudert, 2001). Ontologies are developed to create the planning marketplace comprising demands and offers from different sources (multiple shippers and carriers). Findings A hierarchy ontology for food products. A transporter system ontology. A global ontology that contains all shared concepts used by local ontologies of both shippers and carriers. I-POVES an interoperable model, which facilitates collaboration between carriers and their shippers through its active agents. Practical implications I-POVES is tested on a case study from the TECCAS Poctefa project, comprising transport and food companies from both sides of the Pyrenees (France and Spain). Originality/value There has been much work in the literature on the delivery of products, but very few on the delivery of food products. Work related to delivery of food products focuses mostly on timely delivery for avoiding its wastage. In this paper, constraints related to food products and to environment (pollution and carbon impact) of transport resources are taken into account while planning the delivery.
... Optimization applied to parameter tuning of the agents has been performed with both a single objective by Calvez [6] and multiple objectives by Rogers [7]. Specific instances of the utilization of agent based simulation in conjunction with various optimization approaches have been presented by Deshpande [8] and Gjerdrum et al. [9] for manufacturing scheduling, Sirikipanichkul [10] for freight hub location, Neagu [11] for transport logistics and by Botterud [12] for expansion in electricity markets. It is clear from the wealth of research activity in the application of agent based modeling and simulation that the approach offers a valid means for addressing complex, real world problems. ...
Full-text available
Agent based simulation has successfully been applied to model complex organizational behavior and to improve or optimize aspects of organizational performance. Agents, with intelligence supported through the application of a genetic algorithm are proposed as a means of optimizing the performance of the system being modeled. Local decisions made by agents and other system variables are placed in the genetic encoding. This allows local agents to positively impact high level system performance. A simple, but non trivial, peg game is utilized to introduce the concept. A multiple objective bin packing problem is then solved to demonstrate the potential of the approach in meeting a number of high level goals. The methodology allows not only for a systems level optimization, but also provides data which can be analyzed to determine what constitutes effective agent behavior.
... Distributed agent-based decision systems have provided satisfactory solutions given the nature of the problem, which is dynamic, unpredictable and distributed. An example of this category of solution is the Living Systems Adaptive Transportation Networks (LS/ATN) (Neagu et al. 2006). ...
The logistic sector raises a number of highly challenging problems. Probably one of the most important ones is the shipping planning, i.e. plan the routes that the shippers have to follow to deliver the goods. In this article, we present an artificial intelligence-based solution that has been designed to help a logistic company to improve its routes planning process. In order to achieve this goal, the solution uses the knowledge acquired by the company drivers to propose optimised routes. Hence, the proposed solution gathers the experience of the drivers, processes it and optimises the delivery process. The solution uses data mining to extract knowledge from the company information systems and prepares it for analysis with a case-based reasoning (CBR) algorithm. The CBR obtains critical business intelligence knowledge from the drivers experience that is needed by the planner. The design of the routes is done by a genetic algorithm that, given the processed information, optimises the routes following several objectives, such as minimise the distance or time. Experimentation shows that the proposed approach is able to find routes that improve, on average, the routes made by the human experts.
Full-text available
This paper proposes the use of computational methods of Complex Networks Analysis to augment the capabilities of broker agents involved in multi agent freight transport negotiation. We have developed an experimentation environment that enabled us to obtain compelling arguments suggesting that using our proposed approach, the broker is able to apply more effective negotiation strategies for gaining longer term benefits, than those offered by the standard Iterated Contract Net negotiation approach. The proposed negotiation strategies take effect on the entire population of biding agents and are driven by market inspired purposes like for example breaking monopolies and supporting agents with diverse transportation capabilities.
Agents and agent-based approaches are an active research topis in artificial intelligence and expert systems.
In order to achieve the generalization of inspection system of various industries, an inspection support system based on J2EE architecture, B / S mode is designed and implemented. It can achieve the requirements of most industries through configuration, and the second development is available to achieve the user's specific needs. This paper describes the system composition, the overall structure and the functions of the background management subsystem. Struts, Hibernate frameworks are used in the realization of the background management subsystem.
The purpose of this paper is to present an outlook on the future research for a specific mixed model transportation network referred to as Foliated Transportation Networks (FTN). FTN is thus far a conceptual model that is based on the idea of foliating a direct shipment and a hub-and-spoke structure in order to achieve higher fill rates without an increase in the total traffic work at the same time. The conclusion is that the two principal areas of research are areas of planability (i.e., the ability to in advance and on a sufficient level of detail and precision determine the capacity requirements of the system) and network optimization (i.e., the optimization of the distribution of goods and resources between the different layers of the network).
Conference Paper
Full-text available
Abstract This paper discusses a technological solution to real-time road transportation optimization using a commercial multi-agent based system, LS/ATN, which has been proven through real-world deployment to reduce transportation costs for both small and large fleets in the full and part load business. Subsequent to describing the real-time optimization approach, we discuss how ,the platform is currently evolving to accept live data from vehicles in the fleet in order to improve optimization accuracy. A selection of the predominant pervasive technologies available today for enhancing intelligent route optimization is described.
Full-text available
In pickup and delivery problems vehicles have to transport loads from origins to destinations without transshipment at intermediate locations. In this paper, we discuss several characteristics that distinguish them from standard vehicle routing problems and present a survey of the problem types and solution methods found in the literature.
Conference Paper
Full-text available
This paper describes LS/ATN, Living Systems®Adaptive Transportation Networks, an agent-based solution we have developed to solve transportation problems in the charter business logistics. LS/ATN provides automatic optimization and execution capabilities that extend the existing planning systems accordingly. To describe our solution and analyse its performance, we report on a real case scenario in which transportation requests of a big logistics provider were optimized. Besides describing the agent approach and the LS/ATN features we stress the necessity to integrate such agent system into a real-world IT architecture. Finally, we show that our adaptive solution produces significantly better results in real case scenarios than what achieved with manual optimization of professional dispatchers.
Full-text available
A multiagent approach to designing the transportation domain is presented. The Mars system is described which models cooperative order scheduling within a society of shipping companies. We argue why Distributed Artificial Intelligence (DAI) offers suitable tools to deal with the hard problems in this domain. We present three important instances for DAI techniques that proved useful in the transportation application: cooperation among the agents, task decomposition and task allocation, and decentralised planning. An extension of the contract net protocol for task decomposition and task allocation is presented; we show that it can be used to obtain good initial solutions for complex resource allocation problems. By introducing global information based upon auction protocols, this initial solution can be improved significantly. We demonstrate that the auction mechanism used for schedule optimisation can also be used for implementing dynamic replanning. Experimental results are provided ev...
A heuristic algorithm is described for a time-constrained version of the advance-request, multi-vehicle, many-to-many Dial-A-Ride Problem (DARP). The time constraints consist of upper bounds on: (1) the amount of time by which the pick-up or delivery of a customer can deviate from the desired pick-up or delivery time; (2) the time that a customer can spend riding in a vehicle. The algorithm uses a sequential insertion procedure to assign customers to vehicles and to determine a time schedule of pick-ups and deliveries for each vehicle. A flexible objective function balances the cost of providing service with the customers' preferences for pick-up and delivery times close to those requested, and for short ride times. Computational experience with the algorithm is described, including a run with a real database of 2600 customers and some 20 simultaneously active vehicles. The scenario for the application of the algorithm is also discussed in detail.
Vehicle routing problems (VRP's) involve assigning a fleet of limited capacity service vehicles to service a set of customers. This paper describes an innovative, agent-based approach to solving a real-world vehicle-routing problem embedded in a highly dynamic, unpredictable domain. Most VRP research, and all commercial products for solving VRP's, make a static-world assumption, ignoring the dynamism in the real world. Our system is explicitly designed to address dynamism, and employs an in-time algorithm that quickly finds partial solutions to a problem, and improves these as time allows. Our fundamental innovation is a stochastic improvement mechanism that enables a distributed, agent-based system to achieve highquality solutions in the absence of a centralized dispatcher. This solution-improvement technology overcomes inherent weaknesses in the distributed problem-solving approach that make it difficult to find high-quality solutions to complex optimization problems...
: Dynamic vehicle routing and dispatching refers to a wide range of problems where information on the problem is revealed to the decision maker concurrently with the determination of the solution. These problems have recently emerged as an active area of research due to recent technological advances that allow real-time information to be quickly obtained and processed. There are probably as many variants of these problems as there are real-world applications. After a brief overview of this broad domain, the paper will then focus on problems motivated by courier services and demand responsive transportation systems. These systems typically evolve within a local service area over a relatively short period of time (typically, a few hours), thus putting stringent response time requirements on the decision maker. They are also characterized by a strong routing component: many tasks can be allocated at once to the same vehicle and these tasks must be appropriately sequenced. The ...
Long haul freight transportation
C. T.G. Long haul freight transportation. R.W. Hall, editor, Handbook of Transportation Science, 2002.