Defining Reasonably Foreseeable Parameter Ranges Using Real-World Traffic Data for Scenario-Based Safety Assessment of Automated Vehicles

Abstract

This work was supported by the Ministry of Economy, Trade and Industry of Japan through the SAKURA Project (https://www.sakura-prj.go.jp/). This work involved human subjects or animals in its research. Approval of all ethical and experimental procedures and protocols was granted by the Ethical Committee of the Japan Automobile Research Institute under Application No. 20-014 and 21-017, and performed in line with the Code of Ethics and Conduct published by the Japanese Psychological Association.

Cut-Out Scenario Generation With Reasonability Foreseeable Parameter Range From Real Highway Dataset for Autonomous Vehicle Assessment

“…For example, although the lateral velocity of the cutting-out vehicle appeared to be affected to a lesser extent by the speed range variation compared with other parameters, the distribution indicates a wider range of Vy under the high-speed range than the low-speed range. This result accords with our earlier observations of the lateral velocity in the cut-in scenario [45], which established that a lane-change maneuver that is performed with a lateral velocity less than or equal to 2 m/s is not aggressive (i.e., not time-critical maneuvers). However, these maneuvers can be safety-critical considering the distance to the surrounding vehicle and the relative velocity between vehicles.…”

“…Several techniques have been developed to extract, parameterize, and generate test scenarios for scenario-based assessment; most of them are based on extracting specific scenarios from a real-driving dataset. In our previous research [45], we proposed a methodology to build a Gaussian mixture model, from which representative test cases (that account for multidimensional dependencies) can be generated (see Appendix-A). The previous research focused on cut-in and deceleration scenarios extracted from real-world traffic data recorded by instrumented vehicles and infrastructure cameras under the Safety Assurance KUdos for Reliable Autonomous vehicles (SAKURA) initiative (https://www.sakura-prj.go.jp/).…”

“…At the present, to our best knowledge, there is no specific, unified, method that can comprehensively process real-world data, parameterize them, and quantify the result in a specific safety range applicable for every scenario type. SAKURA research proposes an analytical approach to extract and parameterize scenarios and quantify the resulting parameters with reasonably foreseeable parameter ranges based on risk exposure and acceptance to generate representative test cases [45]. The applicability of the proposed method has been elaborated by processing and comparing two different datasets from Japan and Germany using cut-in and deceleration scenarios [35].…”

“…For extraction of cut-out parameter ranges, the SAKURA dataset was adopted. The dataset was collected with three instrumented vehicles on limited-access highways in Japan [45]. The recordings took place between Nov. 2018 and March 2020 in different weather and traffic conditions.…”

Holistic Perspectives on Safety of Automated Driving Systems - Methods for Provision of Evidence

“…Scenarios can be generated from real data, as suggested in, e.g. [5,6,7,100,101], synthesised based on models, or through expert knowledge, e.g. by using an ontology [102,103].…”