Determination of the Minimum Number of Possible Testing Situations in Autonomous Driving Using Critical Phenomena

Kiss, G.; Bakucz, Peter; Szeghegyi, Ágnes

doi:10.3311/pptr.20767

Cited by 1 publication

(1 citation statement)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The computational burden worsens when traffic simulation is used in conjunction with sophisticated communication models [25]. For example, simulation-based testing of Connected Automated Vehicles (CAVs) require extensive scenarios involving lots of vehicles interacting with each other in various traffic situations [26], [27].…”

Section: Introductionmentioning

confidence: 99%

EGO-Centric, Multi-Scale Co-Simulation to Tackle Large Urban Traffic Scenarios

2023

View full text Add to dashboard Cite

Simulating automotive functions that rely on interaction with other vehicles (e.g., perceptionbased control or algorithms relying on inter-vehicular communication) created a demand for traffic simulation in the automotive field as well. Large-scale traffic simulation can be used to generate long, synthetic drive-cycles for EGO vehicles with realistic traffic. An EGO vehicle is defined as the vehicle the scenario revolves around, presumably running a control algorithm to be tested. On the other hand, simulating an entire district or city with thousands of vehicles present is superfluous and comes with a heavy computational burden while only the vicinity of the EGO vehicle is relevant. On the other hand, major traffic patterns can that could still influence the nearby traffic (e.g., traffic disruptions farther away) but can be simulated with lesser accuracy. Thus, simulation accuracy far from the EGO vehicle can be traded for simulation speed. This paper achieves this trade-off by co-simulating SUMO in microscopic and mesoscopic modes using Libsumo API. Microsimulated traffic is continuously spawned in an EGO-centered sub-network based on traffic states in the mesoscopic simulation. Simulation results in large urban scenarios suggest that the behavior of the EGO vehicle in terms of velocity distribution, headway distribution, and lane changes accurately matches pure microsimulation while simulation speed increased by 3 − 10 times. This result assumes linear time complexity control algorithms with respect to the vehicle number and a single EGO vehicle. Reducing the number of microsimulated vehicles with co-simulation yields even larger simulation speed gains for more computationally complex algorithms. The aggregate (macroscopic) traffic parameters match for both the micro-, meso-, and co-simulated cases. Thus coupling the two simulators does not distort the mesoscopic simulation.

show abstract

Section: Introductionmentioning

confidence: 99%