Nearly all the simulations that use computers are carried out by changing large numbers of parameters or model. This trial-and-error process may be performed several thousand or several tens of thousands of times, which results in an enormous amount of data. An ongoing concern for many researchers is the effort required for managing this information manually and the time required to input the information for later retrieval of data. RIKEN Center for Computational Science (R-CCS) has developed an Organizing Assistant for Comprehensive and Interactive Simulations (OACIS) with an aim to address these problems.
The main feature of OACIS is that it automates much of the common work involved during performing simulations, such as logging in to the remote hosts, creating directories, building execution scripts, sending results, conducting analysis, and managing the information.
It features a versatile and easy-to-use interface that simplifies the use of these functions.
In previous management tools, it was necessary to define a workflow by inserting parameters into the scripts one-by-one, which is no longer required with OACIS. Jobs can be executed simply by selecting the parameters and other settings via internet. Thus, OACIS will allow the essential trial-and-error simulations to be executed more rapidly and in greater quantities in the initial stages of the study, particularly when the conventional procedures have not yet been determined. OACIS includes a function that displays an image of the managed simulation along with the parameters in the form of a correlation diagram. This not only facilitates the execution and management but also facilitates the casual comparison and analysis of data.
Hiroyasu Matsushima, of the National Institute of Advanced Industrial Science and Technology, simulated an evacuation in Kamakura City. In Kamakura, there is a stark contrast between the populations during the day and night. Thus, this simulation was carried out with detailed settings, taking into account the state of congestion and number of evacuees based on the time of evacuation during a disaster. The geographical factors were also considered. The results were statistically analyzed using the design of experiment method. Based on their analysis, a model was derived to provide a highly reliable and effective evacuation guidance.
An evacuation simulation during a disaster at Kamakura City. The evacuation takes place toward the evacuation points designated by the agents represented by the green points on the road.
OACIS played a key role in executing and analyzing this evacuation simulation. Matsushima, who simulated a large number of situations in his research, highly rated the functionality of OACIS as it operates in a browser, allows simple parameter entry, and makes it easy to retrieve, analyze, and manage the test results.
He says that; “I really appreciate its ability of not only executing a large number of simulations but also for managing them consistently. In particular, there is a function to allow the user to visualize the relation between a parameter and results in the form of a graph. This is extremely convenient when I need to confirm the general test trends and to extract results that I am particularly interested in. Another main feature of the OACIS API(*) is that it can be easily extended to different applications. I am using it not only for the social simulations but also to optimize evolutionary computation. In the near future, machine learning will enable the software to handle an even larger number of parameters owing to the recent developments in the field of machine learning.”
(*)API: Application Programming Interface. OACIS not only operates in a browser but also can be used to automate operations in script languages such as Ruby or Python.
Till date, many researchers have independently written scripts and managed them manually to perform computerized simulations. OACIS eliminates the effort required in writing these scripts and thus facilitating error-free computational simulations.
As OACIS speeds up procedures through automation, researchers have more time to concentrate on the important aspects of their research. More simulation attempts can be made by further simplifying the parameter entry, which will enable more situations emerging from more possibilities and more ideas will be simulated.
The greatest strength of OACIS is its versatility. This is an open-source software and can be constructed swiftly in UNIX environment. Thus, it can be applied at any research site where computational simulation is required.
In future, OACIS will ultimately provide more opportunities for simulation and more time for research. As OACIS continues to evolve, the scope of its application will continue to widen.
(2017.12 Interview)