Interaction-Based Task Group Scheduling for a Scalable and Real-Time Self-Driving Simulation

Akihito Kohiga, Kei Hiroi, Takumi Kataoka, Sho Fukaya, Yoichi Shinoda

2023

Abstract

We developed a self-driving test environment based on Virtual Machine (VM) technology. Our test environment enables us to test multiple types of self-driving Artificial Intelligences (AIs) in one simulation and import their self-driving software to an actual car without modification. Our test environment divides a map into several segments and assigns a physical machine to each segment. If a vehicle crosses a border, the VM that embodies the vehicle is moved to another physical machine. However, VM movement causes a load imbalance on the physical machines. We suggest a novel approach to assign high-priority processing to a group of vehicles. The group of vehicles consists of one target vehicle that we would like to test and other vehicles that may interact with the target vehicle. We create such a group by identifying interactions using a ” pre-simulation”. Our approach reduced the processing time of the jobs created by the self-driving AIs by more than 90 percent under an ideal condition . This result indicates that our approach contributes the real-time processing when the CPU was in an overcommitted state.

Download


Paper Citation


in Harvard Style

Kohiga A., Hiroi K., Kataoka T., Fukaya S. and Shinoda Y. (2023). Interaction-Based Task Group Scheduling for a Scalable and Real-Time Self-Driving Simulation. In Proceedings of the 13th International Conference on Simulation and Modeling Methodologies, Technologies and Applications - Volume 1: SIMULTECH; ISBN 978-989-758-668-2, SciTePress, pages 287-294. DOI: 10.5220/0012083800003546


in Bibtex Style

@conference{simultech23,
author={Akihito Kohiga and Kei Hiroi and Takumi Kataoka and Sho Fukaya and Yoichi Shinoda},
title={Interaction-Based Task Group Scheduling for a Scalable and Real-Time Self-Driving Simulation},
booktitle={Proceedings of the 13th International Conference on Simulation and Modeling Methodologies, Technologies and Applications - Volume 1: SIMULTECH},
year={2023},
pages={287-294},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012083800003546},
isbn={978-989-758-668-2},
}


in EndNote Style

TY - CONF

JO - Proceedings of the 13th International Conference on Simulation and Modeling Methodologies, Technologies and Applications - Volume 1: SIMULTECH
TI - Interaction-Based Task Group Scheduling for a Scalable and Real-Time Self-Driving Simulation
SN - 978-989-758-668-2
AU - Kohiga A.
AU - Hiroi K.
AU - Kataoka T.
AU - Fukaya S.
AU - Shinoda Y.
PY - 2023
SP - 287
EP - 294
DO - 10.5220/0012083800003546
PB - SciTePress