Self-exploring autonomous drone: ready to rescue survivors in underground locations that first responders cannot reach

Combined with the latest AI technologies, drones can autonomously navigate through a complex environment using only its onboard sensors such as laser scanners and cameras, searching for survivors trapped in remote underground areas. In the AI Grand Challenge, a premier AI-centered competition hosted by the South Korean government, Team USRG won the top prize of 1 million US Dollars by successfully demonstrating their self-exploring drone technology.

Article | Spring 2020

Recently, there was a well-known incident in Thailand in which a group of teenagers was trapped in an underground cave. The tunnel was long, dark, and narrow, keeping rescuers from reaching the teenagers. People asked, “With all the recent advances in robotics, is there anything that can help these kids to return to their homes safe and sound? Wouldn’t it be possible for drones to fly through the cave and search for the survivors at a moment’s notice?”

The Unmanned Systems Research Group (USRG) in the School of Electrical Engineering of KAIST led by Professor Hyunchul Shim have been actively engaged with various aspects of drone research, ranging from low-level flight control to AI-powered autonomous operations. In particular, USRG has been a forerunner in autonomous drone racing, winning a number of recent events such as Autonomous Drone Racings (ADR) hosted by IROS (1st and 2nd places), Lockheed Martin’s Alpha Pilot in 2019 (3rd place), and Microsoft’s Game of Drone event hosted during NeurIPS in 2019 (2nd place).

The indoor flight capability for drone racing can be extended to problems such as searching for survivors in collapsed buildings. In such cases, accurate mapping of surroundings, path planning, and object detection using AI are very important. Team USRG has constructed a special ruggedized drone platform made of carbon fiber. This drone carries cameras, laser scanners, and inertial measurement sensors along with a lightweight GPU board for running deep learning-based algorithms. A novel flight management system that maps the environment, localizes the drone, and finds a safe path in the unknown environment was developed.

In early 2019, IITP, funded by the Ministry of Science and ICT, announced a new competition called the AI Grand Challenge. In this venue, an autonomous drone flies through a series of obstacles to reach a destination. The track is filled with various objects such as an open window on a wall, vertical poles, a tunnel, trees, and nets. In this competition, Professor Shim’s Team USRG was able to fly through the window, poles, and then the tunnel, winning the top prize of the Minister Award and 1 million US Dollars in prize money as research funding.

In pursuit of fully autonomous drone technology, Team USRG is also teamed up with NASA JPL, MIT, and Caltech to participate in the DARPA Subterranean Challenge, where drones and other autonomous vehicles are required to find various objects such as survivors, backpacks, fire extinguishers, and other objects of interests. As this technology matures, we will be able to use drones to fly in the cave simultaneously to look for any survivors at a moment’s notice.