Sanfilippo, F., & Rañó, I. (2023). Mimicking the Sense of Smell of Search and Rescue (SAR) Dogs: a Bio-inspired Steering Framework for Quadruped Robots. In Jaziar Radianti, Ioannis Dokas, Nicolas Lalone, & Deepak Khazanchi (Eds.), Proceedings of the 20th International ISCRAM Conference (pp. 892–901). Omaha, USA: University of Nebraska at Omaha.
Abstract: Due to their sense of smell and ability to explore areas for missing people, dogs are valuable for search and rescue (SAR). Canines can discover humans under water, under snow, and even beneath crumbling structures because they can smell human scent. Building unmanned autonomous quadruped robots with canine agility is an attractive step to fully replicate the capabilities of dogs. Robots with legs are already capable of mimicking some of the physical traits of dogs, such as the capacity to traverse rough terrain. However, they would need to replicate also the level of sensory perception of a dog to successfully perform SAR operations. To achieve this, a navigation strategy that uses a direct sensor-motor coupling by following the principles of the Braitenberg vehicles is adopted in this work. This paper represents one of the first steps towards the connection of bio-inspired sensor-based steering mechanisms and bio-inspired locomotion for quadruped robots.
|
|
Koki Asami, Shono Fujita, Kei Hiroi, & Michinori Hatayama. (2022). Data Augmentation with Synthesized Damaged Roof Images Generated by GAN. In Rob Grace, & Hossein Baharmand (Eds.), ISCRAM 2022 Conference Proceedings – 19th International Conference on Information Systems for Crisis Response and Management (pp. 256–265). Tarbes, France.
Abstract: The lack of availability of large and diverse labeled datasets is one of the most critical issues in the use of machine learning in disaster prevention. Natural disasters are rare occurrences, which makes it difficult to collect sufficient disaster data for training machine learning models. The imbalance between disaster and non-disaster data affects the performance of machine learning algorithms. This study proposes a generative adversarial network (GAN)- based data augmentation, which generates realistic synthesized disaster data to expand the disaster dataset. The effect of the proposed augmentation was validated in the roof damage rate classification task, which improved the recall score by 11.4% on average for classes with small raw data and a high ratio of conventional augmentations such as rotation of image, and the overall recall score improved by 3.9%.
|
|
Samuel Auclair, Pierre Gehl, Mickael Delatre, Christophe Debray, & Philippe Méresse. (2022). In-depth Analysis of Practitioners' Perceptions about Seismic Early Warning Prior to Aftershocks: The Point of View of the USAR Community. In Rob Grace, & Hossein Baharmand (Eds.), ISCRAM 2022 Conference Proceedings – 19th International Conference on Information Systems for Crisis Response and Management (pp. 740–754). Tarbes, France.
Abstract: Urban Search and Rescue (USAR) teams are particularly exposed to the risk of collapse of buildings due to aftershocks, making concept of earthquake early warning (EEW) particularly interesting. In addition to scientific advances in EEW, it is crucial to understand what are the real expectations and needs of USAR teams, and to what extent EEW solutions could meet them. In this study, we conduct a survey to collect insights from USAR rescuers: it highlights that aftershocks are a major concern for them. In this context, we find that the concept of EEW is very favorably received by the respondents, who consider different types of possible actions upon receipt of an early warning. This study provides a basis for the functional specifications of future solutions of EEW useful to all USAR teams, as well as for the definition of their modalities of engagement on the field.
|
|
Masahiro Watanabe, Yu Ozawa, Kenichi Takahashi, Eri Takane, Tetsuya Kimura, Soichiro Suzuki, et al. (2021). Hardware Design and Tests of SMURF V1 Platform for Searching Survivors in Debris Cones. In Anouck Adrot, Rob Grace, Kathleen Moore, & Christopher W. Zobel (Eds.), ISCRAM 2021 Conference Proceedings – 18th International Conference on Information Systems for Crisis Response and Management (pp. 849–866). Blacksburg, VA (USA): Virginia Tech.
Abstract: When a large-scale disaster such as earthquake occurs, a huge number of victims will be trapped under debris in a wide area. Rescue activities in debris are technically not easy and endanger the first responders. There are several methods for improving safety and efficiency of rescue operation, but their availability is limited to a certain area or short operating time. Our project called CURSOR is developing tools to comprehensively search victims using a large number of ground-based robots entering debris transported by aerial drones. Here we show the development of the exploration robot collecting information with several sensors. The robot system was designed based on the requirements and performance was evaluated by ruggedization tests and mobility tests. No critical problem was found in the durability, and the mobility showed as the same as the ordinary wheel. To improve the mobility, we are planning to apply a proposed unique track mechanism.
|
|
Shalini Priya, Manish Bhanu, Sourav Kumar Dandapat, & Joydeep Chandra. (2021). Mirroring Hierarchical Attention in Adversary for Crisis Task Identification: COVID-19, Hurricane Irma. In Anouck Adrot, Rob Grace, Kathleen Moore, & Christopher W. Zobel (Eds.), ISCRAM 2021 Conference Proceedings – 18th International Conference on Information Systems for Crisis Response and Management (pp. 609–620). Blacksburg, VA (USA): Virginia Tech.
Abstract: A surge of instant local information on social media serves as the first alarming tone of need, supports, damage information, etc. during crisis. Identifying such signals primarily helps in reducing and suppressing the substantial impacts of the outbreak. Existing approaches rely on pre-trained models with huge historic information as well ason domain correlation. Additionally, existing models are often task specific and need auxiliary feature information.Mitigating these limitations, we introduce Mirrored Hierarchical Contextual Attention in Adversary (MHCoA2) model that is capable to operate under varying tasks of different crisis incidents. MHCoA2 provides attention by capturing contextual correlation among words to enhance task identification without relying on auxiliary information.The use of adversarial components and an additional feature extractor in MHCoA2 enhances its capability to achievehigher performance. MHCoA2 reports an improvement of 5-8% in terms of standard metrics on two real worldcrisis incidents over state-of-the-art.
|
|