GPU 병렬처리를 활용한 초대규모 차량 운행기록데이터 맵매칭 성능 개선 연구

초록

Map-matching technology, which is used to correct discrepancies between the measured and actual positions of vehicles in navigation systems, is gaining increased importance due to the rapid development of the transportation industry and related technologies. Consequently, there is a growing need within traffic information systems for high-performance batch processing capabilities that can handle massive volumes of vehicle operation logs and efficiently implement map matching. In this context, the present study reviews the trends and recent achievements in map-matching research, proposes a system architecture for high-performance map matching on large-scale operation logs, and establishes a performance testing environment. Based on this setup, two open-source map-matching programs were selected for performance comparison. To address identified bottlenecks, GPU-based processing techniques were applied in an attempt to improve performance. The results confirmed that utilizing GPUs led to a more than fivefold increase in processing speed compared to CPUs during large-scale batch map-matching tasks. These findings suggest that incorporating GPU technology into traffic information systems can significantly enhance performance for large-volume batch processing of map-matching operations. Future performance gains are expected by further optimizing the source code of GPU-implemented map-matching algorithms for GPU execution and resolving performance degradation issues related to data transfer between the CPU and GPU.

키워드