Wyniki wyszukiwania - BazTech

Ograniczanie wyników

1 Annals of Computer Science and Information Systems

1 2017

Znaleziono wyników: 1

Liczba wyników na stronie

Wyniki wyszukiwania

Sortuj według:

Ogranicz wyniki do:

Collision-free agent migration in spatial simulation

Bowzer Christopher, Phan Benjamin, Cohen Kasey, Fukuda Munehiro

Annals of Computer Science and Information Systems

2017

Vol. 13

65--73

Parallelization of agent-based models (ABMs) is one solution for scaling up their simulation size sufficiently covering more realistic problems. In order to break through memory limitation, some ABM simulators such as RepastHPC and FLAME enabled parallel simulation over a cluster system, (i.e. distributed memory). They visualize to agents remote processors' boundary data as ghost space or facilitate message broadcast among agents, so that agents can still share a full or partial view of their simulation space. Yet, ABMs encounter a parallelization problem where multiple agents may migrate to and thus collide with each other on the same logical coordinates, which should not occur in some applications, (e.g., traffic simulation where two vehicles cannot change to the same lane). Although such collision problems have been addressed algorithmically at a user level where an agent stops before or hops over another agent, moves faster or slower, ticks over time, or cuts coordinates finer, they yet require inter-agent synchronization such as serializing agent migration over all collision-inducing sub-spaces or cells, using a single thread. To facilitate collision-free agent migration more efficiently, we considered two migration algorithms named location-ordered and direction-ordered migration, and implemented them over three ABM simulators: Multi Agent Spatial Simulation (MASS), RepastHPC, and FLAME. This paper discusses about programmability and execution performance among these three simulators in collision-free agent migration.