WPI Develops Bat-Inspired Ultrasound and AI for Drones to Navigate Fog and Smoke
A research team led by Nitin J. Sanket at Worcester Polytechnic Institute (WPI) has developed an innovative ultrasound sensor and artificial intelligence (AI) system. This system allows palm-sized aerial robots to navigate in challenging conditions such as fog and smoke, using limited power and computation, for critical search-and-rescue operations. This significant development, published in Science Robotics, highlights ultrasound as a powerful alternative to existing navigation technologies.
System Inspiration and Functionality
The WPI system draws its inspiration from nature's own master navigators: bats.
Inspired by bats, which navigate dark and damp environments using short chirps and analyzing weak echoes with minimal neural activity, the WPI team's ultrasound-based system uses two tiny sensors and requires little computational power.
This enables small aerial robots to perceive their surroundings and operate independently in hazardous areas.
Autonomous aerial robots typically rely on lidar (light detection and ranging) and radar for navigation. However, these conventional systems are often heavy, power-intensive, and costly. Furthermore, they can be significantly affected by darkness, poor weather, and even propeller noise, which interferes with light-based perception and data analysis.
Drone Customization and Testing
The research team meticulously customized an X-shaped aerial quadrotor drone, approximately 6 inches wide and weighing about 1 pound. This specialized drone was equipped with ultrasound sensors and an acoustic shield, specifically designed to mitigate propeller noise interference.
An advanced AI technique known as deep learning was employed to train the robot's computer. This training focused on analyzing weak ultrasound echo patterns, precisely mimicking how a bat's brain processes sound to understand its environment.
Tests were rigorously conducted in varied environments, including outdoors in a wooded area and indoors with obstacles like transparent plastic or metal poles. Indoor tests also encompassed challenging scenarios featuring darkness, fog, and snow. During these evaluations, the drone operated autonomously for approximately five minutes per flight.
Results and Future Outlook
The robot demonstrated impressive capability, achieving a success rate of 72% to 100% in navigating challenging courses across a total of 180 tests. It did, however, demonstrate less success in avoiding thin objects, such as metal poles and slender tree branches, a limitation attributed to weak signal reflection from these narrow surfaces.
Future work by the team aims to develop even smaller and lighter devices. This advancement could significantly extend flight times for aerial robots utilizing the low-power ultrasound system. Improved flight speeds are also a key goal. Longer flight durations and enhanced agility could substantially boost the effectiveness of these aerial robots in real-world search-and-rescue missions, potentially saving more lives.