But for this to happen, scientists have to be able to deploy robots faster and rescuers need more training on how to use them, the experts said at the Fifth EUCogIII Members Conference, organised by the European Network for the Advancement of Artificial Cognitive Systems, Interaction and Robotics last month (19-20 March) in Germany.
Understanding how robots can assist rescuers, and encouraging scientists to deploy robots more rapidly in real-life situations are two constraints to using them in disaster relief, according to Hartmut Surmann, a researcher at the Fraunhofer Institute for Intelligent Analysis and Information Systems in Germany.
As part of the European Union-funded NIFTi (Natural Human-Robot Cooperation in Dynamic Environments) project, Surmann’s team tested robots during the rescue operation that followed the magnitude 6 earthquake that struck northern Italy in May 2012. Both ground-based robots and aerial drones were used to help firefighters assess if the ruins of the town of Mirandola were safe enough for them to go in to look for survivors.
Scientists worked alongside Dutch, German and Italian firefighters, who were shown how to use the robots and provided feedback to the research team.
“We realised that we need to train firefighters in advance, before a natural disaster occurs. Only by doing this can we make sure that both humans and robots know what they can do for each other and really help in [disaster] situations,” Surmann told SciDev.Net.
The northern Italian earthquakes created a “chaotic and stressful situation”, where rapid decisions were needed, said Surmann, but his team found it difficult to pass on the large amount of data collected by the robots in real time.
In addition, his team learned it was impractical to bring their robots to the disaster area and assemble them there because that could take two to three hours, he says. Instead, the robots need to be ready to be used in as little as ten minutes, to allow the rescue operation itself to start, he adds.
Robots provided images of buildings affected by the quake, said Surmann, adding that these helped rescuers, engineers and architects decide if it was safe to send in human teams.
“Also, if a building has collapsed, we can use a laser scanner to measure how much debris there is and how many trucks are needed to carry that away,” Surmann added.
But rain and successive tremors over the following two months caused more damage to buildings, many of which collapsed, he said, rendering much of the data useless.
“We need to provide information over the whole period of time during which rescue operations are taking place. Maybe we should do a robot survey mission every day, especially if new earthquakes take place,” Surmann says.
A number of papers explaining the results of the project and a list of all the open source software used to control these robots have been uploaded to the project’s website to allow researchers in developing countries to adapt this work to their own circumstances and risk environments.
Surmann says he hopes to apply all this knowledge to the second generation of robots that are due to be tested over the next four years, as part of a new project that started last November.
The Long-Term Human Robot Teaming for Robot-Assisted Disaster Response (TRADR) project builds on the results of NIFTi. It aims to use human-robot collaboration to explore and gather samples from disaster environments. The ultimate goal is for robots and humans to develop their understanding of the disaster area over multiple missions, learning how best to work in the area and how to improve teamwork.
Ricardo Téllez, an artificial intelligence researcher at the Polytechnic University of Catalonia in Spain, said that most robots currently used after natural disasters are controlled by humans in real time and even these face problems when working among concrete and metal debris because their wireless signal then becomes difficult to detect.
“Robots have improved in the last years, but we are still very far from achieving their autonomy. We need to improve this and their perception of their surroundings,” he told SciDev.Net.
Robots work well in controlled environments such as laboratories, but struggle when used somewhere that has been hit by an earthquake or a flood, he added.
“We can teach them to recognise a series of objects under a number of circumstances, but the reality is much more complex than that,” said Téllez. “Robots need to learn by themselves from their environment.”
See below for a NIFTi video of a robot exploring a burning building:
Travel to the conference and accommodation there were paid for by the organisers.