Core Concepts
The core message of this article is to propose a novel multi-robot active target tracking framework that considers the existence of sensing and communication danger zones in the environment. The authors formulate the tracking problem as a nonlinear optimization that balances tracking performance and robot safety, and provide practical approximations to efficiently solve the chance-constrained optimization.
Abstract
This paper presents a multi-robot target tracking framework that accounts for the presence of sensing and communication danger zones in the environment. The authors categorize the adversarial attacks into two types: those that can induce sensor failures, and those that can jam communication channels between robots.
The authors formulate the tracking problem as a nonlinear optimization that aims to minimize the tracking error while ensuring robot safety. They model the danger zones as probabilistic constraints and provide practical approximations to convert the chance-based constraints into deterministic ones, enabling efficient online planning.
The key highlights of the paper include:
- Formulation of the multi-robot target tracking problem with sensing and communication danger zones as a nonlinear optimization problem.
- Design of safe distance conditions towards the two types of danger zones as probabilistic constraints, and practical approximations to solve the computationally challenging problem.
- Thorough evaluations in simulations demonstrating the risk-aware behaviors of robots under different uncertainty levels and risk requirements.
- Hardware experiments validating the robustness and effectiveness of the proposed approach using a team of Crazyflie drones tracking ground robots.
The authors show that the robots exhibit corresponding risk-aware behaviors in response to changes in uncertainty levels and risk requirements. When the uncertainty in the danger source's position increases or the required risk level decreases, the robots maintain a larger distance from the danger zones to ensure their safety, even at the cost of some tracking performance.
Stats
The authors derive the measurement model for the robots, which consists of range and bearing measurements, and use it to compute the uncertainty in the target position estimation.
The probability of sensor failure is estimated through sampling the actual position of the danger source 1000 times from its distribution, and checking if the robot is within the safety clearance for each sample.
The probability of communication jamming is also estimated through sampling the position of the jamming source 1000 times, and checking if the ratio of the robot's distance to the jamming source and its distance to the furthest teammate is below a threshold for each sample.
Quotes
"To address this challenge, we investigate multi-robot target tracking in the adversarial environment considering sensing and communication attacks with uncertainty."
"We design specific strategies to avoid different danger zones and proposed a multi-agent tracking framework under the perilous environment."
"We evaluate the performance of our proposed methods in simulations to demonstrate the ability of robots to adjust their risk-aware behaviors under different levels of environmental uncertainty and risk confidence."