Robots can come in all shapes and sizes, and their form factor can vary depending on the tasks they are designed to perform. For advanced search and rescue, a team from Worcester Polytechnic Institute (WPI) drew inspiration from bats to create palm-sized aerial robots.
The team, led by WPI researcher and assistant professor Nitin J. Sanket of the Department of Robotics Engineering, paired ultrasound sensors with artificial intelligence (AI) to enable the robots to fly through fog, smoke, and other hazardous conditions. The team’s work was featured in Science Robotics and was supported by a grant from the National Science Foundation. By focusing on ultrasound — rather than relying on LiDAR or radar, which can be heavy, power-intensive, and costly — Sanket and the team at WPI developed an X-shaped aerial quadrotor drone about six inches wide that featured an acoustic shield to help damp propeller noise. Using deep learning algorithms, the team trained the aerial drone to analyze weak ultrasound echo patterns in a manner similar to how a bat brain interprets echoes.
Sanket stated that the ultrasound-based system requires only two small sensors and minimal computational power for robots to perceive and analyze their surroundings, enabling successful navigation and executive decision-making. The use of AI enables the bots to operate for longer periods in hazardous environments, compared to today’s modern aerial robots.
The team then conducted tests both indoors, with transparent plastic and metal pole obstacles, and in outdoor wooded areas. A few of the indoor tests were conducted in total darkness, with fog, or snow. The results show that the robot achieved a success rate of 72% to 100% in navigating the obstacle course across 180 tests. The robot was less successful in avoiding obstacles that were thin or slender, such as metal poles or tree branches, due to weakly reflected ultrasound signals.
The overall test results show promise. The next step for the project is to explore smaller, lighter configurations that could extend flight time and increase speed. Sanket told WPI Today that “[b]y creating an ultrasound-based system that needs just two tiny sensors and little computation, we can open up opportunities for small aerial robots to perceive their surroundings, make decisions, and independently operate longer in cluttered, hazardous places where current aerial robots struggle. In a real search-and-rescue mission, a few more seconds of flight time could mean the difference between life and death for a survivor.”
