Robotic dogs do better than humans at identifying nests of invasive fire ants. The robots could be useful in helping eradicate an invasive species that packs a venomous sting.
Robotic dogs can spot invasive fire ant nests more efficiently than human searchers – and they can safely stir up a swarm of aggressive ants by poking their nests, which can help researchers identify the notorious pests.
 |
| Robotic dogs can locate fire ant nests Xin Su | Lanzhou University |
The red imported fire ant (Solenopsis invicta) packs a venomous sting, and causes billions of dollars in ecological and agricultural damage across the world each year. To combat the threat, researchers have been testing how a robotic dog combined with artificial intelligence can help detect the ants’ dirt mound nests. The robot can use its front leg to probe the dirt mound and see if the ants rush out with mandibles and stingers at the ready.
“We ultimately rely on stepping in to actually see the fire ants bursting out to confirm a nest identification,” says Zheng Yan at Lanzhou University in China.
Yan and his colleagues started out by training an open-source AI to visually distinguish fire ant nests from those of other ant species. The training images included nest pictures taken from different angles and in a variety of environmental conditions, including nests hidden in grass, among tall plants and among fallen leaves.
After the training, they used specialised computer hardware carried by a robotic dog to run the AI. The robotic dog is also capable of autonomously exploring a pre-set operational area after being tasked by humans.
In tests, the robotic dog outcompeted three people – who had each received an hour of pest management and identification training – in finding hidden fire ant nests within 300-square-metre nursery garden areas. Two separate field trials were held in China’s Guangdong province and Zhejiang province.
Although both the robot and humans completed the task in less than 10 minutes, the robotic dog detected three times more nests and identified nests more accurately, with a 95 per cent precision rate. But the robotic system had more difficulty identifying smaller nests that had just been founded by the ant colony’s queen.
The robot’s AI currently also cannot distinguish fire ants from other ant species, which is why it confirmed nest identity based on the number of ants and their aggressive behaviour, says Yan. A more capable and expensive robotic dog could also provide better battery life, speed and manoeuvrability in future testing than the Xiaomi CyberDog model used in this case.
Still, such research shows how computer vision and robotics technologies built over the past decade can be combined for use in potential applications such as biological safety and biodiversity conservation, says Deepak Pathak at Carnegie Mellon University in Pennsylvania.
“They are basically using an off-the-shelf vision system and off-the-shelf robotic system,” says Pathak. “But nevertheless, I think this is a very cool application.”
Journal reference:
0 Comments