Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Hypotheses Ego-Tracking for Resilient Navigation

Peter Iwer Hoedt Karstensen, Roberto Galeazzi

TL;DR

This work addresses resilient navigation for RF-based autonomous systems under spoofing and sensor manipulation. It introduces a multi-hypothesis ego-tracking framework paired with a Poisson-binomial windowed detector and a state machine to coordinate operation, diagnosis, and mitigation. A differential-flatness-based path re-planning and NMPC trajectory stabilization enable information gathering and minimal performance loss during attacks. Case studies demonstrate effective detection of biased sensors, maintenance of state estimates, and recovery to nominal operation under persistent spoofing, highlighting practical resilience for GNSS/UWB/5G ISAC-enabled robots. The approach offers a scalable method for isolating malicious measurements and sustaining navigation in adversarial environments, with potential extensions to multi-robot systems and trust-weighted hypothesis management.

Abstract

Autonomous robots relying on radio frequency (RF)-based localization such as global navigation satellite system (GNSS), ultra-wide band (UWB), and 5G integrated sensing and communication (ISAC) are vulnerable to spoofing and sensor manipulation. This paper presents a resilient navigation architecture that combines multi-hypothesis estimation with a Poisson binomial windowed-count detector for anomaly identification and isolation. A state machine coordinates transitions between operation, diagnosis, and mitigation, enabling adaptive response to adversarial conditions. When attacks are detected, trajectory re-planning based on differential flatness allows information-gathering maneuvers minimizing performance loss. Case studies demonstrate effective detection of biased sensors, maintenance of state estimation, and recovery of nominal operation under persistent spoofing attacks

Multi-Hypotheses Ego-Tracking for Resilient Navigation

TL;DR

This work addresses resilient navigation for RF-based autonomous systems under spoofing and sensor manipulation. It introduces a multi-hypothesis ego-tracking framework paired with a Poisson-binomial windowed detector and a state machine to coordinate operation, diagnosis, and mitigation. A differential-flatness-based path re-planning and NMPC trajectory stabilization enable information gathering and minimal performance loss during attacks. Case studies demonstrate effective detection of biased sensors, maintenance of state estimates, and recovery to nominal operation under persistent spoofing, highlighting practical resilience for GNSS/UWB/5G ISAC-enabled robots. The approach offers a scalable method for isolating malicious measurements and sustaining navigation in adversarial environments, with potential extensions to multi-robot systems and trust-weighted hypothesis management.

Abstract

Autonomous robots relying on radio frequency (RF)-based localization such as global navigation satellite system (GNSS), ultra-wide band (UWB), and 5G integrated sensing and communication (ISAC) are vulnerable to spoofing and sensor manipulation. This paper presents a resilient navigation architecture that combines multi-hypothesis estimation with a Poisson binomial windowed-count detector for anomaly identification and isolation. A state machine coordinates transitions between operation, diagnosis, and mitigation, enabling adaptive response to adversarial conditions. When attacks are detected, trajectory re-planning based on differential flatness allows information-gathering maneuvers minimizing performance loss. Case studies demonstrate effective detection of biased sensors, maintenance of state estimation, and recovery of nominal operation under persistent spoofing attacks

Paper Structure

This paper contains 34 sections, 1 theorem, 41 equations, 10 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Contributions
  3. Outline
  4. Related Work
  5. Problem Statement
  6. Notation
  7. Cyber-Physical Resilient System Architecture
  8. Agent & Measurement Model
  9. Measurement Sources
  10. State Machine
  11. Operation and diagnosis state
  12. Mitigation state
  13. Multi hypothesis ego tracking
  14. Hypotheses
  15. Hypotheses estimation
  16. ...and 19 more sections

Key Result

Proposition 1

Let $\hat{\mathbf{z}}\!\sim\!N(\boldsymbol{\mu}_{\hat{z}},\mathbf{P})$ and $\mathbf{z}\!\sim\!N(\boldsymbol{\mu}_z,\mathbf{R})$ with $\mathbb{E}(\boldsymbol{\mu}_{\hat{z}})=\mathbb{E}(\boldsymbol{\mu}_z)$. For the region of inliers $\mathcal{E}_{\mathbf{R}}^{\gamma_{\alpha_\chi}}\!\left(\mathbf{z}\r where $n_{z} = \mathrm{dim}\left(\mathbf{z}\right)$.

Figures (10)

  • Figure 1: Block diagram of the proposed cyber-attack-resilient system architecture.
  • Figure 2: System state machine and transitions.
  • Figure 3: Diagram depicting the operations in a chronological order and how the parameters enter into the algorithm. The vertical arrows contain $\mathcal{H}_k$, $\mathcal{P}_{\mathrm{in},k}^{(\iota)}$, $\left(\mathcal{D}^c\right)^{(\iota)}_{k}$ and $W_k^{(\iota)}$.
  • Figure 4: Graphical visualization of the proof of the outlier probability. The red curve is the measurement density function and the blue curve is the predicted density function.
  • Figure 5: Illustration of how the path re-planned trajectory is computed using a predicted state from the nmpc trajectory stabilization controller. The predicted states are shown in black and the path re-planned trajectory is shown in red. The blue crosses shows the time instances and the blue rectangle shows the view point region.
  • ...and 5 more figures

Theorems & Definitions (2)

  • Proposition 1
  • proof