How to Stop Disagreeing and Start Cooperatingin the Presence of Asymmetric Packet Loss

Abstract

We consider the design of a disagreement correction protocol in multi-vehicle systems. Vehicles broadcast in real-time vital information such as position, direction, speed, acceleration, intention, etc. This information is then used to identify the risks and adapt their trajectory to maintain the highest performance without compromising the safety. To minimize the risk due to the use of inconsistent information, all cooperating vehicles must agree whether to use the exchanged information to operate in a cooperative mode or use the only local information to operate in an autonomous mode. However, since wireless communications are prone to failures, it is impossible to deterministically reach an agreement. Therefore, any protocol will exhibit necessary disagreement periods. In this paper, we investigate whether vehicles can still cooperate despite communication failures even in the scenario where communication is suddenly not available. We present a deterministic protocol that allows all participants to either operate a cooperative mode when vehicles can exchange all the information in a timely manner or operate in autonomous mode when messages are lost. We show formally that the disagreement time is bounded by the time that the communication channel requires to deliver messages and validate our protocol using NS-3 simulations. We explain how the proposed solution can be used in vehicular platooning to attain high performance and still guarantee high safety standards despite communication failures.

Self-stabilizing Multivalued Consensus in Asynchronous Crash-prone Systems

“…More generally, in the context of self-stabilization there are algorithms for group communications [19][20][21], consensus in shared-memory systems [18], wireless communications [27,29,30,[37][38][39][40], software defined networks [8,9], virtual infrastructure for mobile nodes [16,17,45], to name a few.…”

Synchronous Robotic Framework

“…However, it is again a consensus problem and consequently there is no deterministic algorithm that can solve it. However, we can minimize the number of rounds where robots do not agree using the protocol proposed in [8]. In this protocol, when a robot does not receive a message at round r from another robot or when it receives AUT ONOMOUS operation mode from any other robot, then it changes its operation mode to AUT ONOMOUS.…”

“…We consider wireless networks where messages can be dropped at any time which makes it impossible to attain common knowledge deterministically. To overcome the impossibility, we consider the stream consensus protocol proposed by Morales-Ponce et al, [8] that guarantees common knowledge at almost every time with period of disagreement of bounded length.…”