APPLICATION OF AGENT’S PARADIGM TO MANAGE
THE URBAN WASTEWATER SYSTEM
M. Verdaguer, M. Aulinas, P. Escribano and M. Poch
Department of Chemical and Agricultural Engineering and Agrofood Technology, University of Girona
Campus Montilivi, Building PI, Girona, Spain
Laboratory of Environmental and Chemical Engineering, Technological Park of University of Girona
Building Jaume Casademont, Girona, Spain
Keywords: Multi-Agent System application, GAIA methodology, Urban Wastewater System.
Abstract: Urban Wastewater Systems (UWS) are complex and their management is a challenging issue. Each one of
the three principal elements that compose the UWS (i.e. sewer system, urban wastewater treatment plant
and the receiving water) has particular goals to reach. However, the elements of the UWS should be ideally
considered together to perform an integrated management of the UWS. Nevertheless, this approximation,
which seems to be necessary, is not easy. Each one of these elements is in practice managed by a different
entity, which has specific strategies and functions to optimize that sometimes are opposed. In this
communication, a well known agent-oriented methodology –GAIA– is used to model the relations that take
place in the UWS. A prototype is implemented in Java using Repast in order to evaluate the possibilities of
agent-oriented methodologies to model this kind of complex systems.
1 INTRODUCTION
Integrated management of Urban Wastewater
Systems (UWS) constitutes a complex problem.
When analyzing the water quality at river basin
level, several sources of pollution are considered for
their implication in the flow and the water quality of
the receiving media (e.g. treated wastewater, runoff,
rainfall water, etc.). These factors are intertwined
and vary over space and time. They make the system
very complex to model, to represent and to
understand. The quality in the upper waters of the
river can affect down waters. Hence, it is important
to consider these elements as a whole (Erbe and
Schütze 2005; Schmitt and Huber 2006; Fu et al.
2008).
Many other factors, apart from the ones directly
affecting the quality of the receiving water,
intervene in the UWS and have implications in the
water quality at a river basin. As follows, some of
the relevant ones are the population, weather
conditions, industrial activities, wastewater
treatment plants (WWTP) and sewer system
elements.
The flow of wastewater in the UWS considered
in this communication is depicted in Fig. 1. As
shown, the UWS comprise a retention tank that
permits to collect rainfall waters separately. A direct
connection of this tank to the receiving water is also
available, preventing excess of white waters entering
the WWTP during extreme rainfall events. Each
industry is connected to a tank that permits to store
for some time its wastewater. Moreover, special
pollutants can be diverted to a different tank, which
can not discharge into the sewer system.
Each one of these elements fulfils one or several
specific purposes, acting as an autonomous entity,
but also interacting with other elements of the
system. The final quality of the receiving media will
depend on the good performance of these
interactions, which are more than a simple
aggregation of individual actions.
The consideration of the agent’s paradigm and
Multi-Agent Systems (MAS) in this context seems
to be suitable. Agent-oriented approaches are good
in representing the interaction between autonomous
entities (from now agents) that hold specific
individual beliefs but interact with each other in
order to achieve a global goal (Sycara, 1998;
Wooldridge, 2001). A state of the art in agent-based
environmental applications is given in Cortés and
Poch (2008).
497
Verdaguer M., Aulinas M., Escribano P. and Poch M. (2009).
APPLICATION OF AGENT’S PARADIGM TO MANAGE THE URBAN WASTEWATER SYSTEM .
In Proceedings of the International Conference on Agents and Artificial Intelligence, pages 497-500
DOI: 10.5220/0001660504970500
Copyright
c
 SciTePress

This work considers the use of agents' paradigm
as a methodology to describe the relations that take
place in the UWS. In section 2 the multi-agent
system is briefly described, giving an overview of
the agents involved in the system, their roles and the
interaction between them. Section 3 describes the
results of a simulation applied in a specific case
study to manage the inflow at the WWTP. Finally,
section 4 summarizes the conclusions.
Industry 1
Industry 2
Industry ..n
River Upstream
Local reception
basin
River
downstream
Tank 1
Tank 2
Tank n
Toxics T.
Meteo T.
WWTP
Household
Meteo
Industry 1
Industry 2
Industry ..n
River Upstream
Local reception
basin
River
downstream
Tank 1
Tank 2
Tank n
Toxics T.
Meteo T.
WWTP
Household
Meteo
Figure 1: Physical elements of the Urban Wastewater
System (arrows indicate the direction of the water flow in
the system).
2 MULTI-AGENT SYSTEM
Multi-Agent Systems (MAS) are built on the basis
of the agent’s notion. Accordingly, MAS can be
defined as a loosely coupled network of autonomous
and heterogeneous agents that interact to solve
problems that are beyond the individual capabilities
or knowledge of each one (Sycara 1998; Jennings et
al. 1998; Wooldridge 2001).
The agent-based model of the UWS proposed in
this work has been developed using the GAIA
methodology (Zambonelli et al. 2003). In GAIA the
basis of the methodology is the good description of
roles and protocols.
An example of one of the roles is shown in Fig.
2, in which there is a brief description of the role,
together with the protocols, permissions and
responsibilities.
As shown in Fig. 2, liveness and safety
properties are very important in order to ensure
favourable states for the system. During the working
cycle, each of the liveness and safety values will be
compared with the values provided by the sensors.
The execution of the actions will follow a different
path as a function of these values.
ROLE: Discharge of the retention tank of rainfall waters
Liveness:
Value of the level measured in the tank ≤ Maximum
value admissible level in the tank.
Safety
Maximum admissible level in the tank retention not to
overload the WWTP
Description : Discharge of rainfall waters from the retention
tank into the receiving media or into the sewer system
Protocols & Activities:
Check level. Calculate the
difference between the measured value and the admissible
maximum settled in order to prevent WWTP’s overload.
Inform discharge.
Permissions:
Read level sensor
Responsibilities:
Figure 2: Schema for role “Discharge of the retention tank
of rainfall waters”.
2.1 UWS Agents
The MAS developed to describe the UWS elements
and interactions comprises the following agents:
Meteorology Agent: develops tasks of
information management associated with the
collection of meteorological data from sensors,
executes its model, and send the output to the
receiver agent depending on the decision taken.
Household Agent: requests and receives data
related to wastewater from households. Sends this
data, properly treated, to the coordinating agent, and
decides whether to send it or not to the council
agent.
Industry Agent: requests and receives data from
industrial retention tanks. Decides upon industrial
wastewater discharges.
Coordinating Agent: requests and receives
information from WWTP, households, meteorology
and industry agents. Manages the information
received in order to take a decision and prioritize
industrial discharges to the WWTP. It is also in
charge to prevent temporary overflows to the
receiving media.
WWTP Agent: manages data from WWTP
sensors and laboratory. According to this, decides on
its own performance and operation, and sends its
status to the coordinating agent.
Local Reception Basin Agent: corresponds to the
receiving media. It receives values of water quality
and quantity related to the discharge and from the
river upstream. Accordingly, calculates the dilution
capacity of the river at this point. It decides whether
to put restrictions on the parameters for the next
simulation cycle.
Basin Council Agent: coordinates the actions
between different river sections. Take the final
decision with regard to each fluvial section.
ICAART 2009 - International Conference on Agents and Artificial Intelligence
498

Reception of
information
Prioritization
Discharge
authorized
WWTP
capacity
Conflicts?
no
yes
Resolution
yes
no
Possible
Execution
Waiting for the next cycle
Reception of
information
Prioritization
Discharge
authorized
WWTP
capacity
Conflicts?
no
yes
Resolution
yes
no
Possible
Execution
Waiting for the next cycle
Figure 4: Schematic decision tree of the Coordinating Agent.
2.2 Agent’s Communication
and Protocols of Interaction
In the previous sections the type of exchanged
information was described, whereas in this section
the focus is on the flow of this information, that is,
which agents can establish communication between
them. It is important to notice that, whereas the flux
of wastewater is unidirectional (see arrows in Fig.
1), the flux of associated information can be
bidirectional or multidirectional (Fig. 3).
Industry
Agent 1
Coordinating
Agent
Household Agent Meteorology Agent
WWTP
Agent
Basin Council
Agent
Industry
Agent 2
Industry
Agent n
Local Reception
Basin Agent
Industry
Agent 1
Coordinating
Agent
Household Agent Meteorology Agent
WWTP
Agent
Basin Council
Agent
Industry
Agent 2
Industry
Agent n
Local Reception
Basin Agent
Figure 3: Agents’ communication channels.
The accomplishment of the established safety
properties for each protocol is a precondition for its
initialization.
2.3 Agents’ Decision Models
In the case study considered in this work, the key
issue is coordinating the discharges from industry
agents and take the decisions about their
prioritization.
When deciding upon the authorization of an
industrial discharge, first of all, the Coordinating
Agent receives the information from Industry Agents
concerning the flows and loads of industries’
discharges. The discharges are kept in industrial
retention tanks till the final authorization on whether
or not to discharge to the WWTP is taken. The
overall decision tree is depicted in Fig. 4.
The fundamental aim of the Coordinating Agent
is to prevent that the contributions of wastewater
into the WWTP overcome its capacity. For each
cycle, it must be ensure that the combination of
flows and pollutant loads do not overcome the
maximum permitted thresholds.
3 SIMULATION OF THE
MULTI-AGENT SYSTEM
In order to evaluate the use of the agents’ paradigm
to manage the UWS interactions a prototype has
been implemented. The agent-based platform used is
Repast (North et al. 2006). The implementation has
been programmed in Java.
The first step of the simulation process consists
in introducing the safety properties of each agent
and reading sensor data from data bases. These
values are compared and the result of the
comparison is sent to Coordinator Agent.
The second step is the coordination of household
and rainfall waters discharges with industrial
wastewater discharges. Household discharges are
prioritized in front of industrial discharges. The third
step is the coordination of industries in order to
prioritize their discharges. The main criterion is the
assignation of specific time for each industry (as a
function of the distance to the connected WWTP) in
order to make the discharge and to check when the
APPLICATION OF AGENT’S PARADIGM TO MANAGE THE URBAN WASTEWATER SYSTEM
499

values surpass the safety ones. Hence, it is possible
to guarantee that the arriving discharges do not
surpass the WWTP safety values.
In Fig. 5 an example of the kind of results
obtained are presented. In this case study, the
objective is to maintain a value of the wastewater
flow at the input of the WWTP as nearest as possible
to the design flow of the plant, which guarantees its
maximum efficacy. The system is capable to
maintain near constant values at the input of the
plant, optimizing the use of the tanks.
Figure 5: Evolution of WWTP inflow along the time.
4 CONCLUSIONS
In this communication it has been presented a
preliminary prototype of UWS description using the
agent’s paradigm.
The system has been developed following GAIA
methodology. Roles and interactions of agents
considered to model the UWS have been described.
The prototype has been applied to manage the
inflows at WWTP in order to minimize the
variations, despite perturbations of different sources.
Results obtained, show that this approach can
deal efficiently with the description and
management of this kind of systems. Further work
is needed to, in the future, identify some emergent
properties at the level of the integrated Urban
Wastewater System by means of more simulations
and new scenarios.
REFERENCES
Cortés, U. and Poch M., 2008. Advanced Agent-Based
Environmental Management Systems. Whitestein
Series in Software Agent Technologies and
Autonomic Computing. In Press.
Erbe V. and Schütze M. 2005. An integrated modelling
concept for immission-based management of sewer
system, wastewater treatment plant and river. Water
Science and Technology 52(5):95-103.
Fu G., Butler D. and Khu S.T. 2008. Multiple objective
optimal control of integrated urban wastewater
systems. Environmental Modelling and Software
23:225-234.
Jennings, N.R., Sycara, K. and Wooldridge, M., 1998. A
Roadmap of Agent Research and Development.
Autonomous Agents and Multi-Agent Systems, 1, 7-
38. Kluwer Academic Publishers, Boston.
North, M.J., Collier, N. T., Vos J.R., 2006. Experiences
creating three implementations of the Repast Agent
Modeling Toolkit. ACM Transactions on Modeling
and Computer Simulation 16(1):1-25.
Sycara K.P., 1998. Multiagent Systems. AI Magazine
19(2):79-92.
Schmitt T.G. and Huber W.C. 2006. The scope of
integrated modelling: system boundaries, sub-systems,
scales and disciplines. Water Science and Technology
54(6-7):405-413.
Wooldridge, M. and Jennings, N.R., 1995. Intelligent
Agents: Theory and Practice. The Knowledge
Engineering Review 10(2):115-152.
Zambonelli, F., Jennings, N.R. and Wooldridge, M., 2003
Developing Multiagent sSystems: The Gaia
methodology. ACM. Transactions on Software
Engineering and Methodology 12(3):317-370.
ICAART 2009 - International Conference on Agents and Artificial Intelligence
500