Scalable Gas Sensing, Mapping, and Path Planning via Decentralized Hilbert Maps

Abstract

This paper develops a decentralized approach to gas distribution mapping (GDM) and information-driven path planning for large-scale distributed sensing systems. Gas mapping is performed using a probabilistic representation known as a Hilbert map, which formulates the mapping problem as a multi-class classification task and uses kernel logistic regression to train a discriminative classifier online. A novel Hilbert map information fusion method is presented for rapidly merging the information from individual robot maps using limited data communication. A communication strategy that implements data fusion among many robots is also presented for the decentralized computation of GDMs. New entropy-based information-driven path-planning methods are developed and compared to existing approaches, such as particle swarm optimization (PSO) and random walks (RW). Numerical experiments conducted in simulated indoor and outdoor environments show that the information-driven approaches proposed in this paper far outperform other approaches, and avoid mutual collisions in real time.

Multi-Agent Distributed Optimal Control for Tracking Large-Scale Multi-Target Systems in Dynamic Environments

“…3) Optimal Microscopic Control for each Mobile Target of the LSTS: With the knowledge of the optimal evolution of the macroscopic state of the LSTS system (note that at the k-th time step, we can compute the LSTL PDF p * k+1 , corresponding to the next time step), each target must utilize a local microscopic control law that will induce the transition of the LSTL's macroscopic state to p * k+1 while ensuring safety at all times. To compute the microscopic control actions for all the N 2 targets, we utilize the artificial potential field (APF) path planning (centralized) approach proposed in [12], [13], [30]. Although, the focus of this work is to solve the multiagent control problem in a distributed way, we still need to briefly describe how one can solve the microscopic control problem for the targets of the LSTS for completeness of our approach but also in order to have to have a framework for fair comparison between our approach with those proposed in [12], [13], [30] when it comes to the multi-agent motion coordination problem.…”

Adaptive Online Distributed Optimal Control of Very-Large-Scale Robotic Systems

“…Since the Hilbert occupancy map learning is not the key contribution, its implementation details are omitted in this paper. The interested readers are referred to [28], [29] for more details. Based on the updated Hilbert occupancy map h(x, t), the obstacle map function is defined as a binary function, such that…”

“…which is an approximation of distribution velocity defined in (27). Then, the Lagrangian term, L (℘ k , m k , Ĉ * k ), reflects the energy-cost E k in (28).…”

“…The microscopic control inputs can be determined according to the sum of the attractive and repulsive gradients. More details can be found in [28].…”

Indoor Environment Monitoring in Search of Gas Leakage by Mobile Robot

“…Another approach of multiple sensors to discriminate gases can be seen in [7], that develop a system with sensors that together distinguish different concentrations of propane, acetone, and ethanol. In order to detect large scale gas, [8] applies the decentralized Gas Distribution Map (GDM) method. Generating a Hilbert map through probabilistic representations, it addresses the task of finding gas concentration in the multiple classes.…”