Resilient Active Information Acquisition with Teams of Robots

Brent Schlotfeldt, Vasileios Tzoumas, George J. Pappas

Abstract

Emerging applications of collaborative autonomy, such as Multi-Target Tracking, Unknown Map Exploration, and Persistent Surveillance, require robots plan paths to navigate an environment while maximizing the information collected via on-board sensors. In this paper, we consider such information acquisition tasks but in adversarial environments, where attacks may temporarily disable the robots' sensors. We propose the first receding horizon algorithm, aiming for robust and adaptive multi-robot planning against any number of attacks, which we call Resilient Active Information acquisitioN (RAIN). RAIN calls, in an online fashion, a Robust Trajectory Planning (RTP) subroutine which plans attack-robust control inputs over a look-ahead planning horizon. We quantify RTP's performance by bounding its suboptimality. We base our theoretical analysis on notions of curvature introduced in combinatorial optimization. We evaluate RAIN in three information acquisition scenarios: Multi-Target Tracking, Occupancy Grid Mapping, and Persistent Surveillance. The scenarios are simulated in C++ and a Unity-based simulator. In all simulations, RAIN runs in real-time, and exhibits superior performance against a state-of-the-art baseline information acquisition algorithm, even in the presence of a high number of attacks. We also demonstrate RAIN's robustness and effectiveness against varying models of attacks (worst-case and random), as well as, varying replanning rates.

Introduction

Active Information Acquisition (AIA) is the process of purposefully moving robots to maximize sensor data collection, such as mapping unknown buildings or tracking moving targets. While standard AIA algorithms assume perfect sensor operation, real-world adversarial or hazardous environments present significant risks. Sensors can be temporarily blinded by smoke, jamming, or hardware failures. If a team's trajectory plan does not account for these failures, losing key robots can compromise the entire mission.

This paper introduces RAIN (Resilient Active Information acquisitioN), a receding-horizon planning algorithm that generates trajectory plans inherently resilient to a specified number of simultaneous sensor failures. By anticipating worst-case attacks during the window when replanning is unavailable, RAIN ensures the team continues to gather critical information.

Methodology

RAIN formulates the problem as a sequential mixed-integer optimization task to maximize information gathered while minimizing the impact of a worst-case attack disabling up to $\alpha$ robots. The core of the algorithm is the Robust Trajectory Planning (RTP) subroutine, which utilizes a two-step "bait" strategy:

  1. The Bait: RTP identifies the $\alpha$ robots that would individually gather the most information and assumes an intelligent attacker will target them.
  2. The Backup: RTP then plans trajectories for the remaining robots assuming the bait robots' sensors are disabled.

This strategy forces the team to adopt overlapping coverage strategies. The authors provide mathematical proofs bounding the suboptimality of RTP using the concept of "curvature" from submodular combinatorial optimization, ensuring the approximation algorithm maintains performance even in worst-case scenarios.

Experimental Evaluation

The authors evaluated RAIN across three simulated environments: 2D Multi-Target Tracking, 2D Occupancy Grid Mapping, and a high-fidelity 3D Persistent Surveillance scenario built on Unity and ROS using the "Camp Lejeune" environment.

Unity-based simulation environment for persistent surveillance.

Key Results

RAIN was compared against a state-of-the-art, non-resilient baseline across all tasks. In Multi-Target Tracking, RAIN achieved peak Root Mean Square Errors (RMSE) that were 2 to 30 times lower than the baseline, particularly when the number of attacks ($\alpha$) was high. In mapping tasks, RAIN consistently achieved lower map entropy, indicating higher certainty. Even when the baseline was allowed to replan at every time step, RAIN improved landmark observation frequency in surveillance tasks by approximately 24%.

Comparison of Mean and Peak RMSE between Non-Resilient and RAIN algorithms across different scenarios and attack levels.

Discussion and Future Work

The results demonstrate that planning for a highly intelligent, worst-case attacker generalizes well, making the system robust to both targeted attacks and random system failures. Current limitations include a reliance on centralized communication for coordination. Future work will focus on extending this framework to distributed settings where robots must plan resiliently using only local communication with immediate neighbors.

Summary updated

Abstract

Emerging applications of collaborative autonomy, such as Multi-Target Tracking, Unknown Map Exploration, and Persistent Surveillance, require robots plan paths to navigate an environment while maximizing the information collected via on-board sensors. In this paper, we consider such information acquisition tasks but in adversarial environments, where attacks may temporarily disable the robots' sensors. We propose the first receding horizon algorithm, aiming for robust and adaptive multi-robot planning against any number of attacks, which we call Resilient Active Information acquisitioN (RAIN). RAIN calls, in an online fashion, a Robust Trajectory Planning (RTP) subroutine which plans attack-robust control inputs over a look-ahead planning horizon. We quantify RTP's performance by bounding its suboptimality. We base our theoretical analysis on notions of curvature introduced in combinatorial optimization. We evaluate RAIN in three information acquisition scenarios: Multi-Target Tracking, Occupancy Grid Mapping, and Persistent Surveillance. The scenarios are simulated in C++ and a Unity-based simulator. In all simulations, RAIN runs in real-time, and exhibits superior performance against a state-of-the-art baseline information acquisition algorithm, even in the presence of a high number of attacks. We also demonstrate RAIN's robustness and effectiveness against varying models of attacks (worst-case and random), as well as, varying replanning rates.
BibTeX
@article{schlotfeldt2022resilient,
  author = {Brent Schlotfeldt and Vasileios Tzoumas and George J. Pappas},
  title = {{Resilient Active Information Acquisition with Teams of Robots}},
  journal = {IEEE Trans. Robotics},
  year = {2022},
  url = {https://openreview.net/forum?id=c3vb2sfM5X},
}
Figure