Digital Twins for Autonomous Driving: A Comprehensive Implementation and Demonstration

TL;DR

This work advances the application of digital twins to autonomous driving by designing and implementing an end-to-end smart mobility DT that integrates RSUs, edge computing, and cloud-based modeling to support real-time route planning. The system distributes tasks across edge and cloud layers, leveraging a heterogeneous V2X network and HD maps to maintain a coherent cyber-physical representation. Real-world demonstration on a Tokyo Tech campus shows a PDR of $99.53\%$ for DT data and an end-to-end latency of $96.61$ ms, which remains below the $100$ ms 3GPP requirement, validating reliability and timeliness for SSMS and information sharing. The results underscore the practical impact of DT-enabled route planning for safer and more efficient autonomous driving in dynamic traffic scenarios.

Abstract

The concept of a digital twin (DT) plays a pivotal role in the ongoing digital transformation and has achieved significant strides for various wireless applications in recent years. In particular, the field of autonomous vehicles is a domain that is ripe for exploiting the concept of DT. Nevertheless, there are many challenges that include holistic consideration and integration of hardware, software, communication methods, and collaboration of edge/cloud computing. In this paper, an end-to-end (E2E) real-world smart mobility DT is designed and implemented for the purpose of autonomous driving. The proposed system utilizes roadside units (RSUs) and edge computing to capture real-world traffic information, which is then processed in the cloud to create a DT model. This DT model is then exploited to enable route planning services for the autonomous vehicle to avoid heavy traffic. Real-world experimental results show that the system reliability can reach 99.53% while achieving a latency that is 3.36% below the 3GPP recommended value of 100 ms for autonomous driving. These results clearly validate the effectiveness of the system according to practical 3GPP standards for sensor and state map sharing (SSMS) and information sharing.

Figures (6)

• Figure 1: System architecture of smart mobility DT.
• Figure 2: Hardware components and Tokyo Tech. smart mobility field.
• Figure 3: Sequence diagram of route planning service.
• Figure 4: Real-time DT modeling: (a) Scenario #A: no congestion in the road network, (b) Scenario #B: congestion occurs on the straight route
• Figure 5: Autonomous driving operation results: (a) Result #A: driving on default route, (b) Result #B: driving on alternative route