Consensus-based Networked Tracking in Presence of Heterogeneous Time-Delays

Mohammadreza Doostmohammadian, Mohammad Pirani, Usman A. KhanarXiv 2023 · arXiv · OpenReview

TL;DR

This paper presents a novel distributed target tracking system for sensor networks that can tolerate heterogeneous time delays in communication. By employing a linear measurement model and a delay-tolerant consensus algorithm, the system achieves improved accuracy and robustness compared to existing nonlinear approaches, even with significant communication latencies.

What Problem Does This Paper Address?

Accurate tracking of moving targets is crucial in many applications, from surveillance to robotics. Traditional centralized tracking systems can be vulnerable to single points of failure and communication bottlenecks. Distributed tracking, where sensors share information and collectively estimate the target's state, offers a more robust alternative. However, real-world sensor networks often suffer from communication delays, which can be arbitrary and vary between sensors (heterogeneous). Existing distributed tracking methods often struggle with these delays, leading to reduced accuracy or requiring high communication rates. This paper addresses the challenge of developing a distributed tracking system that can effectively handle these heterogeneous time delays without sacrificing performance.

Methods

The authors build upon their previous work by proposing a distributed filtering scheme based on linear Time-Difference-of-Arrival (TDOA) measurements. This linear model is a key innovation, as it avoids the inaccuracies introduced by linearizing nonlinear TDOA measurements, which are dependent on the target's unknown position. The core of their approach is a consensus-based networked filtering algorithm that operates in a single time-scale. This means that at each time step, sensors perform one round of consensus-based information fusion followed by an update using their local measurements.

The algorithm is designed to be delay-tolerant. They model communication delays as bounded and heterogeneous, meaning each link between sensors can have a different, but known, maximum delay. The mathematical framework uses augmented state vectors and matrices to elegantly handle these delays within the consensus update. A crucial aspect is the distributed observability condition, which is less restrictive than local observability and ensures that the entire network can collectively observe the target. The stability of the filtering error is analyzed, and a theorem is provided that establishes a sufficient condition for stable tracking in the presence of these delays.

Datasets

Not discussed in the paper. The simulations utilize synthetic data generated based on the described target dynamics and sensor network model.

Key Insights and Ideas

  • Linear TDOA is a game-changer: Moving away from linearized nonlinear TDOA measurements significantly improves tracking accuracy by avoiding reliance on estimated target positions in the measurement matrix.
  • Single time-scale consensus is efficient: Performing consensus and filtering in a single time step per iteration reduces computational burden and communication overhead compared to double time-scale methods.
  • Heterogeneous delays are manageable: The proposed algorithm is robust to varying communication latencies between different sensor pairs, a common issue in real-world networks.
  • Distributed observability is key: The framework relies on the network as a whole being observable, rather than each individual sensor needing to observe the target locally, which is a more practical assumption for sparse networks.
  • Theoretical guarantees for delay tolerance: The paper provides a theoretical foundation for the stability of the tracking error in the presence of delays, offering confidence in the algorithm's performance.

Key Results

The paper demonstrates through simulations that their proposed distributed filtering scheme achieves bounded steady-state stable estimation errors across multiple sensors, even with significant maximum time delays (up to $\tau_{max} = 20$ time steps). The performance is shown to be superior to linearized nonlinear TDOA methods, as illustrated by a comparison in Figure 2. The simulations also show that the algorithm effectively tracks a maneuvering target in 3D space. While specific quantitative metrics like RMSE reduction percentages are not explicitly stated in a summary table, the figures clearly indicate the effectiveness of the approach under various noise and delay conditions.

Limitations and Future Work

The authors acknowledge that their work can be extended to incorporate localized fault-detection and isolation (FDI) strategies. This would allow individual sensors to identify and reject faulty measurements, further enhancing the system's resilience. They also suggest exploring mobile sensors (like drones or UAVs) as a future direction, which would introduce additional challenges related to maintaining rigid formations for consistent measurement matrices and ensuring targets remain within sensor detection ranges. The paper also hints at the potential for cost-optimal and energy-efficient design of such sensor networks.

Summary updated

Abstract

We propose a distributed (single) target tracking scheme based on networked estimation and consensus algorithms over static sensor networks. The tracking part is based on linear time-difference-of-arrival (TDOA) measurement proposed in our previous works. This paper, in particular, develops delay-tolerant distributed filtering solutions over sparse data-transmission networks. We assume general arbitrary heterogeneous delays at different links. This may occur in many realistic large-scale applications where the data-sharing between different nodes is subject to latency due to communication-resource constraints or large spatially distributed sensor networks. The solution we propose in this work shows improved performance (verified by both theory and simulations) in such scenarios. Another privilege of such distributed schemes is the possibility to add localized fault-detection and isolation (FDI) strategies along with survivable graph-theoretic design, which opens many follow-up venues to this research. To our best knowledge no such delay-tolerant distributed linear algorithm is given in the existing distributed tracking literature.
BibTeX
@misc{doostmohammadian2023consensus,
  author = {Mohammadreza Doostmohammadian and Mohammad Pirani and Usman A. Khan},
  title = {{Consensus-based Networked Tracking in Presence of Heterogeneous Time-Delays}},
  year = {2023},
  eprint = {2302.07511},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2302.07511},
}
Figure