ISSN 1320-0682
|
|
|
|
ISSN 1320-0682 |
| Source: | http://www.complexity.org.au/ci/vol06/gimblett/gimblett.html | Received: | 01/07/1998 | ||
| Vol 6: | Copyright 1998 | Accepted for publication: | 15/10/1998 |
H. Randy Gimblett
Merton T. Richards
Robert M. Itami
Associate Professor
School of Renewable Natural Resources
University of Arizona
Tucson, Arizona. USA
Email: gimblett@nexus.srnr.arizona.edu
WWW: http://nexus.srnr.arizona.edu/~gimblett
Associate Professor
School of Forestry
College of Ecosystem Science
and Management
Northern Arizona University
P.O. 15018
Flagstaff, Arizona. USA 86011-5018
Email: richards@alpine.for.nau.edu
Environmental Planner
Catchment Management and Sustainable Agriculture
Department of Natural Resources and Environment
P.O. Box 500
East Melbourne 3002
Email: Robert.Itami@nre.vic.gov.au
This work introduces and explores the potential of using intelligent agent based modeling and simulation as a tool for examining the complex interactions between recreators and the environment, and interactions between recreators as a means to improving the our understanding of the recreational use of wildland settings. In this research the concept of rule-driven autonomous agents as surrogates for human visitors is introduced. Agents are designed to represent the actions of the individual recreators (hiking, mountain bike riding, and pink jeep tour outfitters). Behavioural rules are derived from visitor surveys and interviews conducted in Broken Arrow Canyon, Arizona. The autonomous agents can be seen to dynamically move over a GIS based model of the Broken Arrow landscape. Line-of-sight calculations determine whether an individual agent is able to `see' other agents and are used as method to record `actual' and `perceived' encounters with other agents. Using agent location maps combined with the underlying GIS data the agents can be observed moving across the landscape, pausing, changing pace, lingering at a view-point etc. A discussion focuses on analysing the resulting behaviours found in these simulations and additionally to explore the influence of alternative trail alignments on recreator movement, congestion and crowding. Some potential future directions for this research are discussed.
Recreational uses of forest lands are among an extensive array of commodities and amenities that are increasingly demanded of forest managers. An in depth understanding of the relationships between recreational and other important uses is essential to effective ecosystem management[42]. Within the human dimension of ecosystem management, recreation and amenity uses of forest lands and the associated benefits of those uses, constitute an important component of management decisions. With the recent interest in the human dimension of ecosystem management, new opportunities are provided to improve upon recreation theory by developing new methods to collect, assess, model and simulate spatially relevant data of recreational use, needs, desires and behaviors in forest settings over time and incorporate these assessment methods into an ecosystem modeling framework.
The recreation assessment of forest lands in ecosystem management requires the interaction of four models: a model of desired recreation settings; a model that expresses the outcomes of recreation behavior in those settings; a model of recreation behavior that predicts the number of users per unit of time in those settings and a model that minimizes conflicts within and between recreation groups from which personal, social, and economic value estimates can be made. The unique character of the recreational use of forest lands both requires this approach and makes it achievable[43].
While technology and it's applications are growing by leaps and bounds, there still seems to be a considerable lack of knowledge and confusion in the area of recreation behavior in forest settings. The spatial orientation and temporal nature of encounters, conflicts, psychological states, experience opportunities and associated benefits between and within groups of recreators is still not well understood. Various authors[17, 33, 27, 50, 34, 45, 40, 32, 11, 16] have all focused on the the nature and extent of conflict between members of specific recreation groups, none to date have examined conflict from a spatial and temporal perspective. Work by [30, 29, 38, 47, 26, 52, 39, 49] have employed a variety of techniques including on-site surveys, recording devices and an experience sampling method[8, 28, 9, 10, 37] to quantify immediate psychological states and desires to get at this issue of dynamic, multimodal experience. Results of studies have varied but shed new light on the nature of the recreation experience. Recent work by [30, 29, 26] are among the few to have successfully used these methodologies to analyze a recreators dynamic experience patterns on-site and found that they varied predictably over the course of an outing and were strongly influenced by site characteristics and site management.
Recreation use and the values humans place on the amenity of public lands constitute an important component of management decisions and yet very little is known about the dynamic nature of the recreation experience, where recreators go in the environments, where conflicts occur and when crowding has a negative effect on the recreation experience and outcomes. The work presented in this paper is guided by the following ideas:
The focus of this paper is to present research undertaken to develop a new form of intelligent decision support and simulation system (IDSS) to assist natural resource managers in assessing and managing dynamic recreation behaviour, social interactions and resulting conflicts in wilderness settings using artificial intelligent agents in the Sedona Ranger District of the Coconino National Forest, Sedona, Arizona. The focus of this research is to utilize simulation techniques for exploring the complex interactions between recreators and the environment, and interactions among recreators as a means to improving the foundation of recreation theory.
Sedona, Arizona has been used in this work because it is typical of many special places which have become ever increasingly popular destinations for local, national and international tourists. The Sedona/Oak creek ecosystem covers nearly 200,000 acres from Sycamore Canyon on the West, to I-17 on the east; from beyond the Mogollan Rim on the north, to Beaverhead Flats and the savanna on the south. Sedona is well know by New Age enthusiasts for its ``Spiritual Vortex". This, combined with the close proximity to the Grand Canyon, Monument Valley and the Navaho and Hopi Indian Reservations, make it an important tourist destination by visitors around the world.
Broken Arrow Canyon near Sedona, Arizona was used to capture visitor use data and demonstrate the prototype software to simulate conflicts between recreation groups over time. The Canyon is popular for day hikers, mountain bikers and people on commercial jeep tours because of the unique spectacular desert scenery of eroded red sandstone. The popularity of this canyon is a problem common to many popular wildland recreation destinations. People are ``loving the place to death" by overuse. This overuse not only has negative impacts on the landscape but also in the quality of the experience people have when they visit. Crowding, conflicts between hikers, mountain bike enthusiasts and jeep tours can create negative experiences in what should be a spectacular and memorable landscape setting, but very little is known about where, why and how these impacts are occurring.
U.S. Forest Service who manages the resource, have been seeking guidance on what actions to take to protect the environment and provide the best possible recreation experience for a increasing diversity of visitors. While conventional survey techniques and public meetings have assisted in acquiring a better understanding of use, the spatial and temporal nature of the recreation experience still remain grossly misunderstood. To date, frameworks such as Recreation Opportunity Spectrum (ROS), Limits of Acceptable Change (LAC) and the growing interest in Benefits-Based Management (BBM) have provided managers with guidelines to assess the recreation opportunities, associated beneficial outcomes and identify where human-induced changes occur and to what degree they are acceptable. However, there no tools currently available for natural resource managers to study and quantify the complex spatial dynamic interactions and resulting impacts of recreational use over time.
The Recreation Behavior Simulator (RBSim), outlined later in this paper was developed to address these complex issues by using computer simulation technology. By simulating human behavior in the context of geographic space, it is possible to study the number and type of interactions that visitors will have within each group and between groups. Agent based modeling techniques are used to instill human-like behavior into artificial agents to explore recreation planning alternatives. If resource managers are to have confidence in the use and results of agent-based simulations it is crucial that the design of the behavioral systems of these agents is grounded in observations of actual human behavior in the physical settings in which they naturally occur. From the recreation behavior, management and conflict assessment perspective, a system of this nature could provide a better understanding of recreation conflict and provide a mechanism to test and assess new assumptions and theories of recreation behaviour (goal interference theory) and beneficial outcomes of experiences. In addition, the results of the simulations yield spatially-explicit, social setting data, could ultimately be used to strengthen and improve the overall predictability and mapability of ROS.
There is a growing interest in the research community for using GIS for modeling spatially-explicit dynamic processes [2, 25, 46, 31, 22, 24, 4, 48]. The use of Individual-Based Models (IBM) is one of the popular approaches to modeling spatially-explicit ecological phenomena. IBMs according to [46] are ``organisms-based models capable of modeling variation among individual and interactions between individuals."
IBMs offer potential for studying complex behavior and human/landscape interactions within a spatial framework. Since spatial information about a phenomena is stored on a georeferenced coordinate system, space within a grid is implicit and relative to the origin of the grid [46]. IBMs offer some basic advantages over current cellular automata and other dynamic spatial modeling approaches for examining spatially explicit phenomena. Since space is continuous and location is explicit in IBMs, individuals can be simulated, independent of the environment. This provides the modeler with the ability to define an individual's behavior, personality traits and interaction rules when encountering other individuals. Computer modeling of most ecological phenomena evolves in simulated time. Since space is continuous and individuals are represented independently, temporal and spatial variability in IBMs can be handled asynchronously (individually updated) versus synchronously (global update) common to most raster-based GIS systems.
One form of individual-based modeling approaches that has recently gained popularity is artificial intelligent agents. Intelligent agents or what is referred to as `agent-oriented programming' is being used to capture behavioral conditions and sets of intercommunication among and within agents that coexist in a environment. Several researchers [35, 41, 36, 51, 12, 5, 4, 24, 3, 1, 44, 19, 18] have taken advantage of the spatially-explicit IBMs, agents and GIS. Each of these researchers uses various techniques to build linkages between the simulation models and GIS.
Only recently have researchers seeking new ways to understand human/environment interactions been exploring simulation as a tool for developing models of human behavior. Recent studies by [15, 20, 7] clearly demonstrate the potential for agent-based modeling techniques to examine human/landscape interactions. These studies utilize a general model of multi-agent simulations based on computation agents that represent individual organisms (or groups of organisms) in a one to one correspondence. These studies seek to understand the process of evolution in the study of ecological and sociological systems. As Drogoul et al.[15] state ``we are interested in the simulation of evolution of complex systems where interactions between several individuals at the micro level are responsible for measurable general situations observed at the macro level. When the situation is too complex to be studied analytically, it is important to be able to recreate an artificial universe in which experiments can be done in a reduced and simulated laboratory where all parameters can be controlled precisely." This work and others examining emergent processes in societies is extremely exciting and is yielding interesting results that would have been hard to obtain without the use of such simulations.
Few studies to date that have explored emergent behavior in individuals or societies have utilized the power of GIS for representing the spatial worlds they reside in and interactions with those worlds. What is surprising is that none of these studies have taken advantage of the power of GIS. Since human behavior is inherently spatial, GIS can provide the worlds that individuals could respond to and function within. Currently, there is no GIS system with IBM capabilities. Dibble[13] states that ``individual-based models do currently exist (Santa FE Institute, Swarm ) and in many ways these systems may offer far deeper insights into human geographic phenomena than any current GIS."
It is clear from an evaluation of current research using spatially-explicit IBMs, reactive agents and GIS to model animal/ landscape interactions that they offer a powerful alternative to previous modeling techniques for exploring emergent, complex, evolutionary processes. The ability to model the differences among groups, local interactions and variability in time and space, as well as the complex, decision making process of an individual, make IBMs an ideal technique for exploring human / landscape interactions.
To represent and simulate an individual's behavior independent of the environment it requires an understanding of their personality traits (which include personal goals/intentions, expectations, length of visit, age etc.) and rules which define how they move and interact with other individuals they encounter and to the physical world or landscape where they are engaging in their favorite activity. This study consists of three phases; first to capture and analyze recreational use data for providing artificial agents with personalities and rules that closely reflect actual recreator behavior; second the development of an GIS-based agent simulator for mimicking recreator behavior overtime; finally testing of the agents in their simulated world under both typical use conditions and imploring alternative management strategies.
An on-site visitor use survey was employed before and after recreational outing over a nine month period to capture data on recreational use, desired beneficial outcomes and conflicting recreational uses in the canyon. Trip motives, expectations, use density, reported contacts and place of encounters have been identified as key factors in a satisfactory recreational experience (Scenic Spectrums Pty Ltd. 1995). A binary measurement was used to solicit response on the type of benefits that were desired (trip motives and expectations) during their visit and to what degree they were able to obtain them. The focus was on recreation as essentially goal-directed behavior [14]. Expectations have been acknowledged as extremely important to goal-oriented approaches to recreation behavior. This measure coincided with Jacob and Schreyer's [34] goal interference definition of conflict. Visitors were asked if a range of benefits were desirable (goals and intentions) and whether they could obtain those benefits over time (goal interference). The benefit types used in this study are well documented in Bruns et al. [6] and Lee and Driver [38], based on research undertaken on other public lands. Desired benefits such as getting away from crowds, reduced stress and physical fitness are strong indicators of recreational satisfaction. Crowding has been shown to be one of the major predictors of user dissatisfaction. The survey was used to identify anything that either made the setting an ideal place for achieving, or interfered with acquiring the desired benefits. So negative detractors and the inability to obtain desired benefits together are used to measure goal interference and conflicts, and imply an inability to obtain desired recreational experiences leading to unsatisfactory outcomes [21, 23].
To derive meaningful recreator profiles of the visitors to Broken Arrow Canyon, cluster analysis first run on the recreation activity respondent data to isolate visitors by activity groups and then later used to aggregate visitors within each group based on desired benefits (goals and intentions). K-Means Cluster analysis allows one to specify the number of clusters desired or in the case of this research to explore the number of significant recreator types that could be found within each activity group. In addition, cross tabulation was used to calculate the frequency of which respondents within the classes derived previously identified the significance of each benefit type. Similarly they were asked to indicate their ability to obtain each of the benefits. This measure provides an indication of how often the respondents loaded on the benefit types by cluster and what particular benefits could not be obtained. This analysis was subsequently used to determine statistically relevant number of agent types within each activity class for subsequently programming artificial intelligent agents with these identical behavior traits. Since it is possible to derive hundreds of agent personality profiles, for purposes of demonstrating the method, this research aggregated agent classes into a reasonable number for final implementation.
Of the (n=1041) visitors sampled, three significant recreation use groups were identified; day-use hikers (n=337), mountain bikers (n=393) and Commercial Jeep outfitters (pink jeep tours) (n=319). While there was an extensive amount of visitor use data collected during the field work, only some is pertinent to this particular paper. For more detailed demographic data see Gimblett[21]. While there could be many combinations of personality traits derived from the visitor data collected, to demonstrate the utility of the agent modeling system the recreator patterns resulting from the cluster analysis were aggregated into two unique types for both the hikers and mountain bikers. These two types are referred to in this work as either a `landscape' or `social' recreator type. Each has significantly different desired benefits of their recreation experience.
Figure 1 illustrates the differences in the two recreator types. A landscape recreator or agent type is one that seeks out landscapes that are physically challenging, avoid crowds subsequently leading to a reduction in stress. This recreator type typically avoids others at all costs. This is evident by the extremely high desire to avoid crowds. In the exit interviews, recreators that were representative of this agent class indicated that they would only stop in locations where there are no other recreators and move as fast as possible along the trails. Physical exercise was a strong motivation in this recreation group and common to both hikers and mountain bikers as can be seen in Figure 2. These recreator types fall within the personal well being and health benefits class as identified in Bruns et al. [6].
Figure 1: The Frequency of Responses of Aggregated Hikers to Desired Benefits
Figure 2: The Frequency of Responses of Aggregated Hikers to Desired Benefits
Table 1: Aggregation of Hiker Agent types for Simulations based on
Benefit Preferences and Age Group
Table 2: Aggregation of Biker Agent types for Simulations based on
Benefit Preferences and Age Group
A social recreator or agent type is more group oriented, one who seeks out those landscapes which are not necessarily physically challenging but tend to build self-confidence, provide more opportunity to learn more about the natural and cultural history of the area and interact with others who share these goals. This is evident in Figures 1 & 2 where their desires to obtain certain types of benefits are not as strong as those representative of landscape recreators. Social agents did not mind encountering other social recreators along the trail in the case of either hikers or mountain biking. During the exit interviews, recreators that represented this class indicated that liked social interaction while engaging in their favorite recreational activity and will spend longer periods of time wandering through the landscape, sitting in special locations, and contemplating life. This was again true of the mountain bikers as seen in Figure 2. Both of these recreator types have been identified over and over again in the applied recreation literature. Tables 1 & 2 outline the aggregation of agent clusters and their associated age groups into agent classes for both Hikers and Mountain Bikers. For more details on the statistical analysis see Gimblett[21].
Figure 3: The Frequency of Responses of Aggregated Hikers Ability to Obtain Benefits
Figure 4: The Frequency of Responses of Aggregated Bikers Ability to Obtain Benefits
Figures 3 & 4 represent the same recreator agent classes as presented in Figures 1 & 2 but in response to the questions asking their ability to obtain the type of benefits that they desired. Figure 3 illustrates there was a reasonable agreement by both groups of recreators that they could satisfy their desired goals, except in the case of the landscape hikers who could not avoid crowds as opposed to the social agents who could or it simply did not matter to them. Figure 4 reports on the mountain biker's ability to obtain their desired benefits differed. The landscape bikers reported an inability to obtain their desire physical challenge benefits and strongly agreed that it was too crowded. Crowds could be the reason they were unable to achieve the physical benefit. The social bikers seemed to agree that most of their desired benefits could be obtained.
While these results are certainly not conclusive, they do provide a method for assessing the goals and intentions of the recreators visiting Broken Arrow Canyon and also provide a measure of how well they were able to meet those goals or satisfied with there recreational experience. While none on the visitors indicated they were totally unsatisfied with their experience, many seemed frustrated with the numbers of encounters they had with other recreators using the canyon.
Rules for providing the simulated agents with social behaviors of human recreators were derived from what respondents told us about their experiences in the surveys, statistical analyses presented earlier and through interviews following their outing. The respondents were all asked to explain the types of behaviors that they exhibited along the trails when encountering other recreators. While the surveys clearly documented that visitors spent a minimum of two hours performing their activities, the desired benefit questions provided the goals and intentions for their visit. The maps provide a clear indication of where they rested, their final destination or where they stopped to view cultural and geologic features. Many of those recreators that subsequently fell into the social agent class indicated that they stopped at all the locations regardless of the numbers of other hikers or bikers that were present and stayed primarily on the appropriate trail for their activity. Some of mountain bikers and hikers who fell into the landscape agent classes indicated in both the survey and later in the exit interviews that they would stop at the cultural and geologic features only if there were a limited number of other recreators present. They also indicated that they would go out of their way to pass others along the trails quickly and avoid them if possible.
For the testing of the prototype agent simulator, Table 3 presents six rules were developed that conformed to what was learned about the intensity of use and interactions of both social and landscape types of recreators using Broken Arrow Canyon. To accommodate the solitude seeking and crowd avoidance desired benefit of the landscape recreator, a rule was developed that prohibits a landscape agent from stopping if there are more than five other agents present at the cultural or geologic features. Landscape agents are programmed to avoid crowds at all costs. They will speed up if they have 50% energy remaining, to pass other agents on the trails if they are within fifty meters and travelling slower. This rule conforms to what the some of the hikers and mountain bikers told us about their trail experiences and adds to the physical challenge that they sought.
To accommodate the needs of the social hiker and biker visitors, the corresponding agents are programmed to hike or ride to areas in the simulated landscape to learn more about natural features and to socialize with other agents. The agents in these classes generally spend at least two hours performing that activity. They have lower desires for extremely challenging physical fitness, but will seek out areas where they can spend time, such as at the cultural or geologic sites. If a social agent encounters a small group or perceives the ability to catch up to another social agent they will increase their speed to do so. They will remain with them throughout the duration of the simulation, unless they expended too much energy and will be forced to slow down and rest. These rules conform to what the hikers and mountain bikers reported about the type of behaviors they exhibit on the trails. Social agents will stop at all cultural or historic features no matter how many other recreators are present.
Since there are a four different recreator age groups being represented in the simulations, they all will move at different rates along the trail, some will run out of energy sooner (older ones) and will be forced to rest. The behavior rules for these social agents (rules 1, 4 & 5) are summarized in Tables 3 and 4.
The behavior of individuals involved in the jeep tours are not as robust as those hiking and mountain biking. Jeeps contain between four and six visitors on each trip. These visitors spend from two to three hours on the jeeps interacting with the driver and other tourists. Jeep agents are modelled as a group of passengers in one jeep which vary their jeep speed according to the topographic conditions. They speed up or slow down according to the degree of slope. Jeep agents conform to only one action rule that defines their behavior and that is Rule 4 (Table 3) to stop at all cultural or geologic features. The time they spend at these features is predetermined and conforms to what the jeep tour drivers typically spend at each location.
Solitude seeking is an important reported desire, goal or expectation in this study. The degree of interference with that goal is related to the number of encounters one has with other recreators. Table 5 illustrates that perceived negative encounters with certain types of recreators have some impact on the experience. The degree of that impact is not yet known. However, from the comments on the survey such as ``seeing too many people", ``too many people", ``too many jeep tours on trails", "seeing jeeps along the trails", suggests that both visual and physical encounters are important measures of the degree to which a goal or desired benefit is interfered with. Based on the fact that both the hikers and mountain bikers reported that they were not able to obtain a crowd avoidance benefit, and the number of negative encounters that were reported, it was hypothesized that visual and physical encounters with other recreators interferes with their ability to acquire those desired benefits. For the purposes of this research, visual encounters within and between agent types are used as a measure of goal interference or the inability to achieve a desired or perceived benefit. The hypothesis being that the higher the degree of crowding induced encounters, the ability of the agent to obtain some of the other desired benefits may decline, lessening the perceived quality of the overall recreational experience.
Table 5: Recreation Conflicts Between and Within Recreator Groups
From a management perspective what is needed is to identify the spatial locations along the trails where there are significant visual encounters. To accomplish this task, each agent keeps track of the number of encounters it has in each cell along the trail and also stores the type and number of visual encounters it has with other agents on other trails. These encounters are summarized, graphed and mapped to examine areas where there are levels of encounters that interfere with the recreator or agent's goal to obtain a desired benefit.
RBSim is a computer program that simulates the behavior of human recreators in high use natural environments. Specifically RBSim uses concepts from recreation research and artificial intelligence (AI) and combines them with geographic information systems (GIS) to produce an integrated system for exploring the interactions between different recreation user groups within geographic space. RBSim joins two computer technologies:
RBSim is experimental at this stage, but demonstrates the potential of combining the two technologies to explore the complex interactions between humans and the environment[22, 24, 21, 23]. The implications of this technology should also be applicable to the study of wildlife populations and other systems where there are complex interactions in the environment.
RBSim uses autonomous agents to simulate recreator behavior. An autonomous agent is a computer simulation that is based on concepts from Artificial Life research. Agent simulations are built using object oriented programming technology. The agents are autonomous because once they are programmed they can move about the landscape like software robots. The agents can gather data from their environment, make decisions from this information and change their behavior according to the situation they find themselves in. Each individual agent has it's own physical mobility, sensory, and cognitive capabilities. This results in actions that echo the behavior of real animals (humans) in the environment.
The process of building an agent is iterative and combines knowledge derived from empirical data with the intuition of the programmer. By continuing to program knowledge and rules into the agent, watching the behavior resulting from these rules and comparing it to what is known about actual behavior, a rich and complex set of behaviors emerge. What is compelling about this type of simulation is that it is impossible to predict the behavior of any single agent in the simulation and by observing the interactions between agents it is possible to draw conclusions that are impossible using any other analytical process.
RBSim is important because until now, there have been no tools for recreation managers and researchers to systematically investigate different recreation management options. Much of the recreation research is based on interviews or surveys, but this information fails to inform the manager/researcher how different management options might affect the overall experience of the user. For example if a new trail is introduced, we might expect that conflicts might be reduced, but to what extent? If we go to a system of scheduling use, what is the impact on the number and frequency of users? More importantly when you have different, conflicting recreation uses, how do different management options increase or decrease the potential conflicts?
None of these questions can be answered using conventional tools. These questions all pivot around issues such as time and space as well as more complex issues such as inter-visibility between two locations. By combining human agent simulations with geographic information systems it is possible to study all these issues simultaneously and with relative simplicity.
RBSim is developed using object oriented programming technology. Figure 5 shows a diagram of the principle components of the simulation program.
Figure 5: Recreation Behavior Simulation System (RBSim)
RBSim is comprised of five major components:
RBSim allows the user to specify the following parameters for the recreation agents:
The recreator class is the most complex object class. It is comprised of a set of properties for age and personality type (landscape or social agent). These properties determine the behavioral rules the agent will follow and the mobility and energy levels of the agent. Behavioral rules relate to how the agent responds to views, and the number and type of other recreation agents. These rules may result in the agent changing speed to overtake or catch up with other agents, slowing down and stopping to rest or spend time at important landscape features or viewpoints.
Hiker and Bike agents also have a system of energy levels programmed. Energy levels and speed of travel are related to the age of the agent. Very young, and older agents will move more slowly than agents in other age groups. In addition, as energy is expended during the simulation, these agents will also need to rest to rebuild energy levels. The length of resting time is determined by the estimated time it takes agents of different age groups to recover. Energy expended is calculated incrementally as the recreator moves along the trail. Uphill travel expends more energy than downhill travel. Resting times are randomized between preset time thresholds to represent variability between real human recreators.
Since much of the perception of crowding is based on visual contact as well as physical contact with other recreationists, a vision system is designed for the agents. The visibility class is a modification of standard GIS line of sight or intervisibility analysis. To reduce the computations required to check for visibility between two points, the visibility class checks for inter-visibility only between points occupied by other recreators referenced in the trail object. The line of sight is calculated taking into account intervening screening effects of topography and vegetation from the GIS databases for elevation and forest cover.
Trails are specified for each agent, for each run. The trails are stored in the trail object which is constructed as a linear list of cells derived from a grid based GIS. For each trail cell the distance from the trail head, the elevation, and landscape features associated with the cell is stored. During the simulation run the trail object also stores the number recreators in each cell. This data structure is designed to minimize the computing time for agent navigation through the trail system. All agents of the same class (hiker, biker or jeep) share the same trail object. The trail object therefore acts as a ``collective memory" for the agents in that respective class. Each agent can reference the trail object to determine the location of other agents on the trail and to determine the trail conditions. As the agent moves from one cell to the next, it de-references its location from the last cell (by subtracting one from the recreator count field for that cell) and references its location in the next cell. Since the hikers, bikers and jeeps follow different trails a unique trail object is created for each recreation type. To test management alternatives for new trails, the user may specify different trail files for each simulation run.
The RBsim runtime simulation engine runs in discrete time steps. At each time step in the simulation, each recreator class (hikers, bikers and jeeps) is evaluated to determine if a new instance (agent object) of that class should be created. For each class of recreator a timer is set which begins incrementing from the start of the simulation run and is reset to zero each time a new recreator agent is generated. In the model parameterization the minimum and maximum times between agents is specified. A random time is generated between the minimum and maximum time each time a new agent is generated. A new agent of the respective class will be generated once the timer reaches the randomly generated time.
The new agent object is generated as an instance of the generic recreator class. When the agent is created, properties are set for age, personality, and agent type. These properties are set based on a randomly generated number (between 0 and 1) which sets the probability for each property. For instance, if 25 percent of the biker agents are of the landscape personality type and 75 percent are of the social personality type, then if the random number is between 0 and .25 the simulation engine will generate a landscape bike agent. If the number is greater than .25 and less than or equal to one, the simulation engine will generate a social bike agent. This same strategy applies to the age distribution as well.
Recreator Agents of the hiker, biker and jeep types are placed in collections for each type. The simulation engine then tracks each agent in each collection. Since the simulation engine is running on a synchronous clock, the order in which the agents are executed will affect consequences such as crowding and visibility. In order to avoid order effects from executing agent movement in a set sequence, the sequence is randomized within each collection for each iteration of the simulation. Each agent has a single method called ``Move" which triggers the execution of the internal rules, energy levels and mobility for that agent. Once the agent has completed execution of all its behaviors for that time step, the runtime simulation engine then executes the move method for the next agent in the randomized list for that iteration. This process continues in a loop until either all agents have completed their journey or the maximum time set for the simulation run is reached.
At the conclusion of the simulation each trail object writes its contents to the output file object. RBsim then returns control to the user.
The synthetic world that the simulated recreators utilize is a
georeferenced, raster database consisting of 513 rows
522
columns, each cell 10 meters square. The database consists of
topography, vegetation, adjacent primary and secondary roads, existing
and proposed trails, trail head, jeep staging area, significant
geologic features and scenic stops. These geographic themes were
deemed important for this work, but many more could be incorporated as
the sophistication of the modeling increases.
The approach taken in this research was to provide each agent with spatial analytical capabilities that is imperative to them processing information necessary for functioning in the simulated worlds. Each agent is provided with the ability to calculate distance or proximity to other agents and significant features in the landscape. Each calculates the percentage of slope from the topographic map and whether it is going up or down hill and in turn speeds up or slows down accordingly. They utilize neighborhood functions to identify trail cells or the location of significant geologic features and scenic stops. Most importantly each agent has visual capabilities for detecting other agents, how far away they are and whether they can or cannot be seen. This algorithm uses forest cover and topography as constraints to detecting other agents.
In order to test some of the ideas and concepts presented in this research and to determine the efficiency of the simulation system in identifying conflicting recreation behavior, a set of experiments were constructed. During the interview and survey phase of this research, visitors to Broken Arrow Canyon were asked in addition to the information already discussed, to record the month, day and time that they entered the canyon.
Table 6 presents a statistical summary of the visits of those recreators sampled over the duration of the study. As can be seen, peak times throughout the canyon are weekends. Over 40% of hikers and bikers frequent the canyon during these time periods, while 40% visit in the morning, with 60% in the afternoon. Week days according to our sampling were the highest for Jeep tours into the canyon. In order to test RBSim, many simulation runs were undertaken mimicking various peak and off-use times to examine the dynamic interactions or recreators and resulting visual and physical conflicts. This study reports on one of those experiments, a mid weekday (Wednesday) which typically is not a peak use period but contains a moderate number of hikers and bikers and relatively low jeep usage. The schedule of use during that typical day is reported in Table 7.
In order to demonstrate a potential management action such as restricting biking use on a heavily used trail, two alternative bike trails were substituted for the original and the simulations rerun to evaluate the differences in recreational use and resulting perceived conflicts from all recreators perspectives (See Figure 6).
Figure 6: Runtime Simulation Interface with both the Original and Alternative Trail Layouts for Broken Arrow Study
Table 7 is an example of mixed recreational use along the trails. Figures 7 illustrate the intensity of hiker encounters with other agents from the hiking trails. Figure 7 illustrates a significant number of encounters with both other hikers and bikers, hikers versus other hikers and bikers and jeeps. Encounters with bikers is high from the beginning of the simulation, peaks at Chicken Point and is chaotic until completing the journey. Hikers, on the other hand, peak at Chicken Point and then remain consistently high thereafter. What is of interest is that where the encounters with hikers peak, biker encounters drop off and visa versa.
Figure 7: Graphed Results of Hiker Encounters with Other Agents from along Hiking Trail
Figure 8: Graphed Results of Biker Encounters with Other Agents from along Biking Trail
As summarized in Table 5, over 40% of the negative encounters that occur to hikers are with jeeps and 30% with other hikers or bikers. It is interesting that even with the number of hikers, bikers and jeeps included in this simulation, that there are very few encounters with jeeps. On the other hand, the high amount of conflicts with bikers and other hikers may have a detrimental effect on the recreation experience. But because the place, time and duration of encounters that occur between biker and hiker agents are not consistent, this may tend reduce to accumulative impact of the encounters on those hiking.
Figures 8 illustrate biker encounters with hikers, jeeps and other bikers from along the biking trail. The patterns are similar to those found in Figure 7 except bike encounters steadily increase throughout the life of the simulation, but are not as high as in Figure 7. Biker's encounters with hikers are more sporadic than is outlined in Figure 7 dropping off at the end of the simulation. Like above, there are virtually no encounters with jeeps.
Figure 9: Graphed Results of Jeep Encounters with Other Agents from along Jeep Trail
Figure 10: Graphed Results of Hiker Encounters with Other Agents from along Hiking Trail
Figure 9 illustrate a high number of jeep encounters with hikers and bikers and only a minimal number of encounters with other jeeps. The encounters occurring with hikers and bikers are concentrated around Chicken and Submarine Rock during the last half of the trip with virtually no encounters occurring for the first and last quarter of the simulation.
In summary it appears that with the increased number of hikers and bikers in the canyon that encounters with bikers are the most dominant impact. There are very few encounters with jeeps throughout all the simulations in experiment 4.
As illustrated in Figure 10, selecting alternative bike trails can have a major impact on the number of encounters that occur along the trails. It can be seen in Figure 10 that when alternative bike route 1 is used in the simulations that the number of biker encounters that the hikers will have decreases significantly to the point that they are negligible after Chicken Point. When compared to Figure 7 and summarized in Table 8, by altering the trail layout the mean number of encounters has dropped by two thirds and the maximum number of encounters by half.
Figure 11: Graphed Results of Biker Encounters with Other Agents from along Biking Trail
Figure 12: Graphed Results of Jeep Encounters with Other Agents from along Jeep Trail
In Figure 11 the number of visual encounters with other recreators that bikers will have when using the alternative bike route reveals a dramatic decline in both hikers and jeeps, but a steady increase in number of bikers. In fact, an evaluation of Table 8 illustrates that visual encounters with hikers declines to one fifth of those that occurred in Figure 8, with the same number of hikers still using the trails. This strongly suggests that by using the alternative trail, the distribution of hikers and bikers within the canyon is more conducive to minimizing conflicts.
Figures 10 illustrate a significant number of encounters with jeeps from both bikers and hikers in the canyon. As in Figures 8 & 9, encounters with other agents declined. Of significance are the encounters with hikers and jeeps. But interestingly enough, increasing the number of bikers from eleven to twenty seven has little effect on the mean number of encounters that occur, but does effect the maximum. In other words while the number of encounters remains the same, the encounters are more evenly dispersed along the trail, rather than peaking at specific locations. From a management perspective, if the objective is to disperse the impacts of encounters over time and reduce high impact areas of conflict, then this alternative bike route would offer a solution to this problem.
The results of this research illustrate however that the conflicts most often reported are from bikers having negative encounters with hikers. While jeeps are certainly considered to have a fairly high level of impact on both hikers and bikers, they are not as strong a determinant of a negative recreational experience as anticipated. Bikers and hikers continually clash in the canyons. What is of interest however is how often and where these encounters occur. An examination of the results of the agent simulation runs illustrates that bikers most frequently clash with other bikers. While bikers may have more encounters with other bikers, as reported in the survey, they do not see them as detracting from their experiences.
The agent simulations seem to be an excellent method for modeling recreator encounters and ultimately conflicts. While statistical results of the survey used in this study provide an indication of the average number of encounters (viewed as negative detractors), the agent simulations provide a dynamic view of these encounters and identify the spatially explicit locations where they occur. The effect of these encounters on the overall recreational experience is still unknown. However, this simulation environment provides a way to test and evaluate many scenarios of recreational use. While the survey provided a quantitative measure of the recreational experience, the simulation environment provides a dynamic, spatial representation of use and provides the added benefit of collecting and storing data on encounters over time. Both these data can be evaluated using conventional statistical techniques and compared to explain commonalities and differences.
Of interest in this research, and showing the power of using simulation, is the impact of alternative routes on recreator encounters. An examination of the biker trail alternatives as suggested by the respondents to the survey, illustrate the importance of a well thought out trail design on recreational encounters. As can be seen in this research both alternative trail designs significantly reduce the number of encounters with other recreators. In fact, from before the turn around point to the completion of a biker's journey, they literally have no visual contact with any other recreator type. If hikers do have an accumulated negative effect on a bikers experience then it is clear that these alternative routes would alleviate this problem. This situation is identical when assessing encounters using the alternative jeep trail layout. Biker and hiker encounters with jeeps virtually disappear for over fifty percent of the journey. This is a substantial decline in encounters considering the significant number reported by the respondents. It is clear that the simulation environment can assist in evaluating existing and proposed trails in an attempt to minimize encounters and conflicts which ultimately lead to a decline in recreational experience.
What is imperative to emphasize in this work is that simulation of any dynamic behavior cannot be accomplished without such techniques as developed in this research. Simulation using personality traits and behavioral rules synthesized from human recreators provide a forum to evaluate and test a diversity of recreator use densities over time. These alternatives can be used to develop new facilities along the trails, and to redirect trail use to maximize user satisfaction while minimizing impact. Being capable of seeing the agents interacting under a variety of constraints can assist the manager in acquiring a better understanding of how human recreators use and interact on public lands.
This research has taken the first step forward to make linkages between GIS, Multiagent systems and Recreation Behavior Modeling. While this research has not directly dealt with goal interference theory, it does use it as a foundation for behavior modeling. It is assumed that perceived benefits, and obtaining or maintaining those benefits, directly correlates with the goal interference. Encounter and/or subsequent conflicts are the main cause of goal interference. The landscape recreator agent developed in this work was programmed specifically to avoid interference and when threatened passed other agents or avoided stopping at scenic lookouts. This technique allows one to reduce the amount of goal interference, while maximizing benefits. More than that, it allows one to experiment with artificial recreators to realistically determine thresholds of goal interference and devise management strategies to reduce it. This is one of the advantages of using simulation and the power of such multiagent environments.
This research advances our knowledge and understanding or natural resource assessment and intelligent simulation systems in the following ways: Extends the theoretical foundation of recreation and behavior by exploring the concept of benefits-based management for measuring desirable and obtainable benefits of leisure and assessing spatially-explicit visual and physical encounters among recreators in Broken Arrow Canyon; Extends the knowledge-base of the development, calibration and use of intentional, multi-autonomous agent systems in GIS represented worlds; Develops an entirely new form of intelligent decision support system (IDSS) to assist natural resource managers in assessing and managing human use of natural areas which could be easily extended into a number of other areas such as assessing impacts on wildlife habitats; Expands the existing capabilities of visual operators found in GIS for providing all mobile agents with visual capabilities. This vision system is used for controlling agent movement, goal-seeking, determining locations and distances of potentially conflicting agents and could be easily modified for identifying significant landscape features; Utilizes conventional social science survey techniques with automated field methodologies for calibrating agent movement; Develops a user friendly, parameterized interface for experimenting with alternative trail layouts and a diversity of agent configurations under a variety of conditions.
Much work can be undertaken to improve the predictability and reliability of the modeling framework. To expand our understanding of the dynamic physiological and psychological experience patterns, sampling methods could be used. Dynamic experience patterns can be empirically measured including such factors as visual acuity, focus of attention, mood, psychological benefits, coping strategies, norms of behavior, and physiological changes at strategic locations within a stratified set of landscape settings found throughout the study site. A methodology employing these techniques that provided a way for the visitor to stop, record and photograph landscapes of importance would provide valuable information and lead to improved understanding of the dynamics of recreation experience. It is important however to ensure that wherever and whenever the visitor records such information, that their explicit location is captured as well so as to be able to link these changes to physiographic settings.
To improve the modeling of social interactions in a physical environment it is imperative that a more thorough understanding be acquired on how humans translate information from the environment into meaningful actions. Human-like agent simulations are no different. Once the spatial information is communicated to an artificial agent it must then be translated from its objective form into the symbolic and cognitive framework from which affective human responses are derived. This area of research needs considerable attention, but will provide meaningful outcomes.
We wish to thank the USDA Forest Service, Rocky Mountain Forest and Range Experiment Station and the Coconino National Forest for their assistance in facilitating this research effort. We also wish to thank Dr. B.L. Driver of the Rocky Mountain Station for his helpful review and oversight of this project. This research was supported in part by funds provided by the Rocky Mountain Forest and Range Experiment Station, Forest Service, U.S. Department of Agriculture.
Note: Instructions for obtaining a free copy of RBSim can be downloaded http://nexus.srnr.arizona.edu/~gimblett/rbsim.html
 |
EMail Contact: Complexity International Editor
