SIP-adus Workshop on Connected and Automated Driving Systems Session : Impact assessment SIP-PROJECT : Preliminary examination of traffic accident simulation to evaluate the benefits of safety systems for the reduction of traffic accidents Osamu Takatori Japan Automobile Research Institute November 18, 2014 1 Project Goal 【National Goal】 •Reduce the number of casualty in the traffic accidents 【Project Goal】 •Develop multi agent simulation tool to evaluate benefit when active safety system and automated driving system are introduced. 【Project Goal in 2014】 •Clarify requirement of the simulation tool developed in the project. 2 Utilization of Simulation Basic development cycle of active safety systems ①Accident statistical analysis ⑨Verification in the real world Society level ②Accident investigation Individual accident level ⑧Benefit estimation ③Driving response Driver response level ⑦Closed loop evaluation investigation （Human factor） ④Set of the target function System level ⑥Open loop evaluation ⑤Develop the active safety system 3 Advantage of simulation Comparison of normal cycle with cycle using simulation in the process of ⑧～⑨ Process①～⑧ Normal cycle Cycle using simulation ⑧Benefit estimation～ ⑨Verification in the real world Around decades Next cycle Next cycle Shorten development cycle Simulation tool leads to Efficient and Effective development. 4 Objective and Work Items in 2014 •Clarify requirement of new simulation tool developed in the project. (1)Investigation of the multi agent simulations using driver model. ①Review of the existing multi agent simulations ②Requirement of driver models ③Applicability of other areas (2)Investigation of requirement on new simulation tool 5 Accident Types ★What accident pattern is focused on? From traffic accident statistics, we will focus on at least three accidents pattern. ①“Rear-end accident” ②“Crossing pedestrian-car accident” ③“accident causing lane departure” 車両単独 工作物 2% 衝突 2% その他の 車両相互 16% 右折時 8% 横断中 横断中 5% 以外 4% 正面衝突 2% 交通事故件数 追突 36% 出会い頭 25% Rear-end accident 36% その他の 車両単独 路外逸脱 4% Fatalities (N=4,278) 横断中 25% 5% 工作物 衝突 15% その他の 車両相互 6% 死亡事故件数 横断中 以外 10% 右折時 6% 出会い頭 13% 追突 6% 正面衝突 10% Crossing pedestrian - car accident 25% 14 12 10 8 6 4 2 0 Fatality rate (%) 12.2 死亡事故率(%) 5.9 3.1 3.0 1.6 0.1 0.4 0.5 0.3 1.6 横 横断 断 中 中 正 以外 面 衝 突 追 出 突 会 そ い の 他 右 頭 の 折 車 時 工 両相 作 互 物 そ の 路 衝 他 外 突 の 逸 車 脱 両 単 独 Traffic accidents in 2013 (N=629,021) その他の Accident causing lane departure 6 Composition of Models In order to reconstruct 3 types of accidents in the simulation, it is necessary to have at least 4 components. Relationship between each model Environment information Environment ・Signal ・Lane etc... Driver Most important but difficult to model ・response time, ・brake work etc... Pedestrian ・walking speed ・initial position ・ etc... Vehicle Safety system ・spec etc... 7 Driver Models ★What data is needed for making driver model？ • In order to develop more accurate simulation, appropriate driver model parameter is necessary(e.g. driver reaction time, brake work etc...). • It is essential to acquire the driver response data by experiments. Driving simulator Vehicle simulator Driver response (reaction time, brake work...) 8 Review of Existing Simulations ★ What simulations are reviewed? • Investigation of the existing multi agent simulation tools, in terms of model construction, functional capability, and extendibility. Various multi agent simulation ASSTREET ・ASSTREET(TOYOTA) ・SIM for ACAT（HONDA，DRI） ・ASSESS(NTSEL) ・MATES(The University of Tokyo） ・SUMO(DLR:Germany） ・Prescan(TNO/Tass International） 9 Review of Existing Simulations ★ How the prospective simulation is selected? • Advantage and disadvantage for each simulation tool are evaluated as simulation requirements. Then, the most prospective simulation is selected. Requirement for the simulation Algorithm of perception・recognition, decision making and control Driver model Normal driving situation Accident avoidance situation Driver error Implementation of individual difference ・・・・・・ Degree of freedom for vehicle dynamics Setup of vehicle specification Setup of automated drive and active safety system ・・・・・・ Road （road geometry、road facilities、traffic lane...） Time zone（day-and-night...） Weather (wet-and-dry...) ・・・・・・ Algorithm of perception・recognition, decision making and control Implementation of individual difference ・・・・・・ Total Evaluation Tool A Tool B Tool C ・・・・・ ◎ ◎ ○ ・ ◎ ◎ ○ ◎ ・ ◎ ◎ ◎ △ ○ ○ ・ ◎ △ △ ◎ × △ ・ ◎ ○ ・ ・ ・ ・ ・ ・ ・ ◎ ○ △ ・ × ○ ・ △ △ ・ × ○ ・ ・ ・ △ △ × ・ ○ ・ ◎ △ ・ ○ × ・ △ ・ ・ ・ • TheVehicle simulation in this project will be developed based model ◎ ○ ○ ・ ・ ・ ・ ・ on the prospective simulation. Environment model Pedestrian model 10 Requirements of Simulation ★What function of the simulation tool should have? Requirements of the simulation tool ○ Driver model：Perception・Recognition, Decision making and Control model Driver error(oversight, Decision making・ Control miss） etc. ○ Vehicle model：Degree of freedom, Automated driving, Active safety system ○ Environment model：Road shape, Traffic lane, Facilities and Other conditions (e.g. Weather, Road surface, Day-and-night) ○ Pedestrian model：Perception・Recognition, Decision making and Action model (Walking speed and walking route) etc... 11 Applicability to Other Areas • Multi agent simulation tools have applicability to not only estimation of traffic accidents but also other area by adding functions (e.g. traffic jam，environment burden). • In this project, extendibility and versatility to other areas are also investigated. Estimation of traffic jam Estimation of traffic accidents + Driver response to ・Traffic sign ・ITS system + Estimation of environment burden ・Gas pedal work ・Engine output ・Gas consumption 12 Summary • We aim at developing the simulation which can carry out exactly benefit estimation of active safety system and automated driving system efficiently through this project. • In promotion of development and spread for active safety technology and automated driving technology, we are sure that a simulation contributes. • Taking advantage of a pioneer's knowledge, it will become a better simulation. 13 Thank you for your attention 14
© Copyright 2024 ExpyDoc
