To better understand the dynamics in hazardous environments, gas distribution mapping aims to map the gas concentration levels of a specified area precisely. Sampling is typically carried out in a spatially sparse manner, either with a mobile robot or a sensor network and concentration values between known data points have to be interpolated. In this paper, we investigate sequential deep learning models that are able to map the gas distribution based on a multiple time step input from a sensor network. We propose a novel hybrid convolutional LSTM - transpose convolutional structure that we train with synthetic gas distribution data. Our results show that learning the spatial and temporal correlation of gas plume patterns outperforms a non-sequential neural network model.
Gas distribution mapping is important to have an accurate understanding of gas concentration levels in hazardous environments. A major problem is that in-situ gas sensors are only able to measure concentrations at their specific location. The gas distribution in-between the sampling locations must therefore be modeled. In this research, we interpret the task of spatial interpolation between sparsely distributed sensors as a task of enhancing an image's resolution, namely super-resolution. Because autoencoders are proven to perform well for this super-resolution task, we trained a convolutional encoder-decoder neural network to map the gas distribution over a spatially sparse sensor network. Due to the difficulty to collect real-world gas distribution data and missing ground truth, we used synthetic data generated with a gas distribution simulator for training and evaluation of the model. Our results show that the neural network was able to learn the behavior of gas plumes and outperforms simpler interpolation techniques.
Simulating a Gas Source Localization Algorithm with Gas Dispersion Produced by Recorded Outdoor Wind
(2023)
This paper reports the use of the first gas dispersion simulator capable of introducing large wind fluctuations into simulations. The proposed simulator enables testing of a modification made to a gas source localization algorithm in a realistic scenario in order to study how the change affects it. Gas source localization in an outdoor environment is a challenging task mainly due to the complexity of the gas spread caused by the unpredictable nature of constantly changing wind. Therefore, a novel use of outdoor wind in developing a gas source localization system by simulation is presented in this paper. To consider the characteristic of sudden but large and unpredictable changes in wind direction, we propose to use recorded outdoor wind to simulate a realistic outdoor gas dispersion which has been done for the first time to the best of our knowledge. With the use of this simulator, we have tested a modification to a mobile robot-based gas source localization algorithm. Multiple simulations of the modified and the original particle filter-based algorithm have been done to study the effect of the tested modification. The results showed that a small difference in the algorithm can greatly impact the results. From this study, we show that the use of simulation consisting of the necessary traits to evaluate outdoor gas source localization, has the potential to accelerate the development of a reliable localization system.
This poster reports an improved outdoor gas source localization algorithm, showing promising results under simulations with a realistic gas plume. In the scenario assumed in this work, an unmanned ground vehicle searches for the location of a gas source in an open outdoor field. The simulated gas plume in this work consists of frequent and large meandering as a real outdoor plume. The source location is estimated using gas and wind measurements obtained from sensors fixed on a mobile platform. When a gas puff is detected, its source is likely to be in the upwind direction. Particle filter-based algorithms have been proposed in previous work to estimate the most likely source location from multiple gas detection events. In this work, the particle weight update function has been modified by adopting a 2D Gaussian plume model, to improve the accuracy in estimating the likelihood of the source location. We have evaluated the performance of this algorithm combined with an estimation-based route planning algorithm. The simulator uses wind data recorded outdoors to calculate transport of gas puffs allowing a formation of a gas plume containing large meandering, due to real fluctuations of outdoor wind. Simulations of the improved particle filter with the estimation-based route planning algorithm have yielded more accurate, stable and time-efficient results than the pre-modified version.