When Distributed Consensus Meets Wireless Connected Autonomous Systems: A Review and a DAG-Based Approach

The connected and autonomous systems (CAS) and auto-driving era is coming into our life. To support CAS applications such as AI-driven decision-making and blockchain-based smart data management platform, data and message exchange/dissemination is a fundamental element. The distributed message broadcast and forward protocols in CAS, such as vehicular ad hoc networks (VANET), can suffer from significant message loss and uncertain transmission delay, and faulty nodes might disseminate fake messages to confuse the network. Therefore, the consensus mechanism is essential in CAS with distributed structure to guarantee correct nodes agree on the same parameter and reach consistency. However, due to the wireless nature of CAS, traditional consensus cannot be directly deployed. In addition, with the development of intelligence in machines, traditional consensus might be inefficient for future CAS. This article reviews several existing consensus mechanisms, including average/maximum/minimum estimation consensus mechanisms that apply on quantity, Byzantine fault-tolerant consensus for request, state machine replication (SMR) and blockchain, as well as their implementations in CAS. To deploy wireless-adapted consensus, we propose a Directed Acyclic Graph (DAG)-based message structure to build a non-equivocation data dissemination protocol for CAS, which has resilience against message loss and unpredictable forwarding latency, and has the potential to solve the first step of future intelligent consensus. Finally, we enhance this protocol by developing a two-dimension DAG-based strategy to achieve partial order for blockchain and total order for the distributed service model SMR, to show its potential of integrating into different consensus protocols.