CHAPTER 2. BACKGROUND LITERATURE:
2.3 The Framework of Collective Problem-Solving
The literature review presented in Section 2.1 and Section 2.2 explained that the environment
has influential contributions in the strategic mechanisms of collective problem-solving in
social insects and multi-agent systems. Based on the literature definitions of the environment,
and the knowledge of the environment roles and responsibilities, the author now presents a
framework of collective problem-solving. The framework establishes an interdisciplinary
knowledge of the environment proposing in a comprehensive form that the influences of
environment contribute to the strategies of collective problem-solving.
First of all, the author explains her understanding of the environment in an interdisciplinary
context, describing and analysing the influences of the environment that contribute to the
strategies and mechanisms of collective problem-solving. The knowledge of environmental
influences is established on the understanding of collective problem-solving in social insects
and multi-agent systems. After that, the author illustrates the framework in Figure 2.8.
Table 2.3 below compares the definitions, the roles and responsibilities of environment in the
collective problem-solving of social insects and multi-agent systems. In the literature, the
environment is defined in biology and in multi-agent systems, where descriptions are given to
the various aspects of the environment. The environment roles and responsibilities in multi-
agent systems are reflected in the relationships between insects and the environment. Such a
reflection indicates that the influences of the environment contribute to the collective
52
Table 2.3 The Environment in Insect Behaviour and In Multi-Agent Systems
Literature In Social Insects In Multi-Agent Systems
The Environment is defined as:
A physical structure that provides chemical and biological
conditions of the region in which the colony requires to live. (1997)
1. Defined in various application-domains by describing the application-specific characteristics of environment
(Rao, Georgeff et al. 1992)
2. Environment = (State, Process) (Van Dyke Parunak 1997)
3. Described using a container function and is defined as “a space E” with a volume (Ferber 1999)
4. “a generic environment programme” illustrates the basic relationship between agents and their environment
(Russell and Norvig 2003)
5. Environment has structures and the nature of mediation
(Odell, Parunak et al. 2003)
6. Environment is an exploitable element in developing multi-agent systems (Danny Weyns 2005)
53
Structures the insect colony
Provides resources and living conditions
Maintains dynamics through stimulating configurations
Is locally observable and accessible to insects
Defines rules for the colony
Structures the MAS
Embeds resources and services
Maintains dynamics
Locally observable and accessible
Defines rules for the MAS
Environment is: Environment Roles:
A medium of coordination (through template, self-organisation and stimulating configurations)
A means for communication (through template and modified environment)
Task-defined environment
A medium of coordination
A means for communication
Application-specific environment
54 The environment is a physical structure. It provides resources and conditions for collective
problem-solving. The environment is an essential element in enhancing effective problem-
solving through the processes of self-organisation, emergent coordination and stigmergic
communication. The environment has effects on the strategies and mechanisms of problem-
solving, which are presented to be: application-specific characteristic properties, stigmergic
communications via the modifications of environment, as well as the effectiveness of
interaction between insects/agents and the environment is altered by the nature of substrate.
In particular, the environment responsibilities and roles outline the influences of environment
on the individual and coordinated behaviour of social insects and multi-agent systems.
The 6 definitions represent the main understanding of environment in multi-agent systems
over the last decade. None of them has officially defined the term, and instead, researchers
intended to describe the various aspects of environment of their own interests. For instance,
Rao et al. (Rao, Georgeff et al. 1992) describe the characteristic properties of the
environment of different application domains; Parunak (Van Dyke Parunak 1997) underlines
the dynamic nature of the environment by assigning values to its states and processes; Ferber
(Ferber 1999) uses a container function to specify the space and volume of an environment;
Russell and Norvig (Russell and Norvig 2003) present the interactive relationship between
agents and the environment; Odell et al. (Odell, Parunak et al. 2003) are also interested in the
structure of environment and describe it to be physical, communicative and social.
In multi-agent systems, the knowledge of environment has been systematically developed.
However, the inspiration of insect colonies in agent-based systems, especially the roles of
environment in collective problem-solving, has been rather implicitly recognised. Because of
its involvement in enabling flexible and robust performance of emergent systems, it is
55 problem-solving. To accomplish this, a framework of collective problem-solving is
constructed to present the interdisciplinary knowledge of the influences of environment in
social insects and multi-agent systems.
Figure 2.8 shows the framework of collective problem-solving, which describes the
interdisciplinary knowledge of the environment through illustrating its characteristic
properties and related mechanisms of problem-solving, in the emergent behaviour of social
insects and the collective problem-solving of multi-agent systems. This framework provides a
theoretical ground for the research work presented here on cooperative search of Unmanned
Aerial Vehicles (UAVs). Learning from the problem-solving mechanisms of social insects
and related applications in multi-agent systems, a set of swarm-inspired mechanisms of
collective problem-solving are designed and deployed to accomplish the cooperative search
problem of UAVs.
56
Figure 2.8 The Framework of Collective Problem-Solving
Environment Roles and Responsibilities
Social Insects Multi-Agent Systems
Influences
Swarm-Inspired Search Strategy Collective Problem-Solving
Pheromone
Stigmergy
Food Distribution and Abundance
Defines Individual Rules of Behaviour
Environment Structure, Resources and Services
Environment Dynamics
Locally Observable and Accessible to Agents
Defines Rules of All Entities in the Multi-Agent Systems
Represented by Represented by Represents
Maintains
Targets,
Resources and/or Structure are
Transform to
Natural Environment Application Environment
Has Has
Achievement and Maintenance Tasks
A Medium of Coordination A Means for Communication
Specifies
Is Is
Template
Self-Organisation
Represents the characteristics and conditions of the environment
Represents the problem specifications and requirements
57 The review so far has explored a series of similarities between insects and multi-agent
systems in collective problem-solving, however, the two also differ from each other in some
aspects. For instance, it is well-known that individual ants and termites are very basic and
simple beings. The idea of mimicking their behaviour in multi-agent systems is to solve
complex problems through the cooperation and coordination of simple individuals with
limited capabilities (Van Dyke Parunak 1997). However, real-time applications of multi-
agent systems have encountered concerns regarding the level of intelligence each individual
agent should be designed with. Regarding their relationships with the environment, agents
can accomplish problem-solving processes with either cognitions or reactions (Ferber 1999).
Cognitive agents have prior knowledge that enables them to solve certain problems by
themselves and to also anticipate future events intentionally; whereas reactive agents simply
respond to the modifications of the environment and have no prior knowledge of the problem.
The former exhibit higher level of intelligence that demands more comprehensive design of
the multi-agent system, and the latter intend to solve problems through emergent
phenomenon.
Despite the emergent behavioural patterns, individual insects present synthesised
characteristics of cognitive agents and reactive agents. Real-world applications consist of
sophisticated tasks that cannot be accomplished by purely cognitive agents or purely reactive
agents. Therefore, it is important to identify a synthesised form of cognition and reaction for
each agent according to the goals and objectives of different problems. A variety of studies
have been carried out to investigate design methodologies and infrastructures that fulfil the
demands of solving complex problems (Parunak, Brueckner et al. 2003; Sierra, Rodriguez-
Aguilar et al. 2004; Park and Sugumaran 2005; Moya and Tolk 2007; Weyns, Omicini et al.
58