Active Sensing Strategy: Multi-Modal, Multi-Robot Source Localization and Mapping in Real-World Settings with Fixed One-Way Switching

TL;DR

The paper addresses gas distribution mapping in real-world, cluttered environments using mobile robots. It introduces a state-machine that switches between budget-constrained coverage path planning (CPP) with flexible formations and swarm-based active sensing (AS), guided by a Gaussian-field gas model and collaborative particle filtering. Key contributions include the first application of CPP to initialize a gas-field map, a collaborative entropy-driven AS tailored for cluttered environments, and a multi-robot switching framework validated through hardware-in-the-loop and real-time experiments, achieving faster convergence, higher accuracy, and collision-free operation. The reported results show a $43\%$ reduction in turnaround time and a $50\%$ increase in estimation accuracy relative to single-mode baselines, underscoring practical gains for robust multi-robot environmental sensing.

Abstract

This paper introduces a state-machine model for a multi-modal, multi-robot environmental sensing algorithm tailored to dynamic real-world settings. The algorithm uniquely combines two exploration strategies for gas source localization and mapping: (1) an initial exploration phase using multi-robot coverage path planning with variable formations for early gas field indication; and (2) a subsequent active sensing phase employing multi-robot swarms for precise field estimation. The state machine governs the transition between these two phases. During exploration, a coverage path maximizes the visited area while measuring gas concentration and estimating the initial gas field at predefined sample times. In the active sensing phase, mobile robots in a swarm collaborate to select the next measurement point, ensuring coordinated and efficient sensing. System validation involves hardware-in-the-loop experiments and real-time tests with a radio source emulating a gas field. The approach is benchmarked against state-of-the-art single-mode active sensing and gas source localization techniques. Evaluation highlights the multi-modal switching approach's ability to expedite convergence, navigate obstacles in dynamic environments, and significantly enhance gas source location accuracy. The findings show a 43% reduction in turnaround time, a 50% increase in estimation accuracy, and improved robustness of multi-robot environmental sensing in cluttered scenarios without collisions, surpassing the performance of conventional active sensing strategies.

Figures (17)

• Figure 1: Even Distribution of 16 Gaussian basis functions on the gas mapping area.
• Figure 2: State machine diagrams: (a) general model for continuous, multi-modal switching (b) specific switching model studied in this paper.
• Figure 3: A demonstration of the CPP algorithm. The blue area is covered by the multi-robot coverage path, while the white area is missed. The black rectangular box represents an obstacle cell. Finally, the red-lined rectangular box describes a real obstacle.
• Figure 4: Selection of the next measurement position (highlighted by a red circle) by a swarm of three Jackals (represented as green circles). Image extracted from our base station Graphical User Interface during a live trial.
• Figure 5: Flow chart of the proposed switching exploration algorithm.