Hands-On Field Operational Test Dataset of a Multi-Controller CPS: A Modeled Case Study on Autonomous Driving

Abstract

Cyber-physical systems (CPSs) observe operational environments and continuously decide on actions to achieve goals. In many cases, complex CPSs have several goals to satisfy simultaneously. To develop a CPS with multiple goals, engineers can create various control systems, each contributing to a goal based on a system-of-systems (SoS) engineering perspective. Engineers then conduct field operational tests (FOTs) to collect data for analyzing and optimizing the control systems. However, uncertainties in the physical environment and the emergent behavior of multiple controllers present several challenges in conducting FOTs. We have hands-on experience in performing massive FOTs of a multi-controller CPS to realize engineering challenges in the FOTs. We modeled and developed an autonomous robot vehicle consisting of a lane-keeping system and an adaptive cruise control system. To analyze and optimize autonomous driving, we conducted FOTs of 125 possible configurations of the control systems, each 50 times. This paper presents 1) the model, software, and hardware implementation manuals of our case study on autonomous driving, 2) an FOT log dataset obtained from about 100 hours of driving and its analysis results, 3) research challenges emerging in the multi-controller CPS FOT learned from our hands-on experience, and 4) possible applications of our dataset for future research.