Distributed Observer Design for Tracking Platoon of Connected and Autonomous Vehicles

TL;DR

This paper studies the platoon of connected and autonomous vehicles (CAV) and proposes a distributed observer to track the state of the CAV dynamics and shows the structure of the Kronecker matrix product of the system dynamics and the adjacency matrix of the CAV network.

Abstract

Intelligent transportation systems (ITS) aim to advance innovative strategies relating to different modes of transport, traffic management, and autonomous vehicles. This paper studies the platoon of connected and autonomous vehicles (CAV) and proposes a distributed observer to track the state of the CAV dynamics. First, we model the CAV dynamics via an LTI interconnected system. Then, a consensus-based strategy is proposed to infer the state of the CAV dynamics based on local information exchange over the communication network of vehicles. A linear-matrix-inequality (LMI) technique is adopted for the block-diagonal observer gain design such that this gain is associated in a distributed way and locally to every vehicle. The distributed observer error dynamics is then shown to follow the structure of the Kronecker matrix product of the system dynamics and the adjacency matrix of the CAV network. The notions of survivable network design and redundant observer scheme are further discussed in the paper to address resilience to link and node failure. Finally, we verify our theoretical contributions via numerical simulations.

Figures (5)

• Figure 1: This figure shows the platoon of connected and autonomous vehicles communicating over a wireless network to share information.
• Figure 2: This figure shows a $3$-node/link-connected CAV network. This network preserves connectivity after the failure of up to $3$ nodes or $3$ links, and therefore, the distributed observer remains practical even in the failure of $3$ vehicle nodes or $3$ communication channels.
• Figure 3: This figure shows the mean-square error at every vehicle under the proposed iterative distributed observer \ref{['eq_p']}-\ref{['eq_m']}. As it is clear from the figure the error dynamics is Schur stable with some steady-state residual due to Gaussian measurement noise.
• Figure 4: This figure shows the distributed observer error at every vehicle under the proposed iterative \ref{['eq_p']}-\ref{['eq_m']} and reduced network connectivity, which is Schur stable.
• Figure 5: This figure shows the error at remaining $3$ vehicles under the proposed iterative distributed observer \ref{['eq_p']}-\ref{['eq_m']} after failure of the $3$rd vehicle. The Schur stability of the MSE shows the resilience of the proposed observer to node failure.