Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Decentralized Input and State Estimation for Multi-agent System with Dynamic Topology and Heterogeneous Sensor Network

Zida Wu, Ankur Mehta

TL;DR

The paper tackles decentralized estimation for multi-agent systems subject to unknown inputs, dynamic topology, and heterogeneous sensors, proposing DISKF which preserves privacy by exchanging only estimates and covariances. It combines input fusion via Covariance Intersection (CI) with a rapid weighting scheme and state fusion via information-filter decomposition, effectively achieving unbiased, minimum-variance estimates comparable to fully informed central filters while requiring only a single communication iteration. A time-window observation mechanism mitigates bias from intermittent observations, and a diffusion step helps correct errors in dynamic topologies; extensive simulations show DISKF outperforms baselines in 1-hop and dynamic scenarios and matches global-information performance in all-to-all settings. The approach has practical impact for cooperative tracking, ad-hoc sensor networks, and geophysical monitoring, where privacy, robustness, and real-time operation are critical, and it naturally extends to large spatial-temporal deployments while preserving local information.

Abstract

A crucial challenge in decentralized systems is state estimation in the presence of unknown inputs, particularly within heterogeneous sensor networks with dynamic topologies. While numerous consensus algorithms have been introduced, they often require extensive information exchange or multiple communication iterations to ensure estimation accuracy. This paper proposes an efficient algorithm that achieves an unbiased and optimal solution comparable to filters with full information about other agents. This is accomplished through the use of information filter decomposition and the fusion of inputs via covariance intersection. Our method requires only a single communication iteration for exchanging individual estimates between agents, instead of multiple rounds of information exchange, thus preserving agents' privacy by avoiding the sharing of explicit observations and system equations. Furthermore, to address the challenges posed by dynamic communication topologies, we propose two practical strategies to handle issues arising from intermittent observations and incomplete state estimation, thereby enhancing the robustness and accuracy of the estimation process. Experiments and ablation studies conducted in both stationary and dynamic environments demonstrate the superiority of our algorithm over other baselines. Notably, it performs as well as, or even better than, algorithms that have a global view of all neighbors.

Decentralized Input and State Estimation for Multi-agent System with Dynamic Topology and Heterogeneous Sensor Network

TL;DR

The paper tackles decentralized estimation for multi-agent systems subject to unknown inputs, dynamic topology, and heterogeneous sensors, proposing DISKF which preserves privacy by exchanging only estimates and covariances. It combines input fusion via Covariance Intersection (CI) with a rapid weighting scheme and state fusion via information-filter decomposition, effectively achieving unbiased, minimum-variance estimates comparable to fully informed central filters while requiring only a single communication iteration. A time-window observation mechanism mitigates bias from intermittent observations, and a diffusion step helps correct errors in dynamic topologies; extensive simulations show DISKF outperforms baselines in 1-hop and dynamic scenarios and matches global-information performance in all-to-all settings. The approach has practical impact for cooperative tracking, ad-hoc sensor networks, and geophysical monitoring, where privacy, robustness, and real-time operation are critical, and it naturally extends to large spatial-temporal deployments while preserving local information.

Abstract

A crucial challenge in decentralized systems is state estimation in the presence of unknown inputs, particularly within heterogeneous sensor networks with dynamic topologies. While numerous consensus algorithms have been introduced, they often require extensive information exchange or multiple communication iterations to ensure estimation accuracy. This paper proposes an efficient algorithm that achieves an unbiased and optimal solution comparable to filters with full information about other agents. This is accomplished through the use of information filter decomposition and the fusion of inputs via covariance intersection. Our method requires only a single communication iteration for exchanging individual estimates between agents, instead of multiple rounds of information exchange, thus preserving agents' privacy by avoiding the sharing of explicit observations and system equations. Furthermore, to address the challenges posed by dynamic communication topologies, we propose two practical strategies to handle issues arising from intermittent observations and incomplete state estimation, thereby enhancing the robustness and accuracy of the estimation process. Experiments and ablation studies conducted in both stationary and dynamic environments demonstrate the superiority of our algorithm over other baselines. Notably, it performs as well as, or even better than, algorithms that have a global view of all neighbors.
Paper Structure (9 sections, 24 equations, 12 figures, 2 tables, 1 algorithm)

This paper contains 9 sections, 24 equations, 12 figures, 2 tables, 1 algorithm.

Table of Contents

  1. INTRODUCTION
  2. Problem Formulation
  3. System description
  4. Input fusion via Covariance Intersection
  5. State fusion via information filter decomposition
  6. Simulation and Experiment Result
  7. Experiment Setup
  8. Result
  9. Conclusion

Figures (12)

  • Figure 1: An example application of a decentralized estimation system is geophysical environment detection, which involves a multi-layered, heterogeneous sensor network with a dynamic topology. Gauges within the river reach offer localized measurements, albeit within a confined spatial range. Satellites afford broader spatial observations but lack real-time monitoring capabilities. Drones complement the observations by providing additional data points.
  • Figure 2: Scenario 1: Four static agents are located in the four quadrants. Two of them can observe only the y-coordinate of the target, while another two are limited to observing only the x-coordinate. Each agent is tasked with observing the target exclusively when it enters its respective quadrant. The communication is established according to three distinct typologies.
  • Figure 3: Scenario 2: 4 static (red) agents and 5 mobile (green) agents compose a dynamic system. Observation of the target by an agent is contingent on sharing the same quadrant with the target, with each agent limited to observing a single target dimension. The trajectories of the mobile agents are delineated by green dashed lines. Each agent possesses a pre-defined communication range, enabling information exchange solely when the communication ranges of two agents intersect.
  • Figure 4: The time window review deals with the abrupt bias caused by input estimation facing intermittent measurements. When the sensor abruptly measures the target, within the period of the time window review, only do state estimation without input estimation.
  • Figure 5: The input estimation with 1-hop communication. The index of nodes corresponds to Fig.\ref{['demo_1']}. Missing estimation means the corresponding algorithm can't recover the input from the data they observe or exchange.
  • ...and 7 more figures