Mihir Garimella took his inspiration from fruit flies. After a family trip to India some rotten bananas attracted hordes of flies. The flying insects were easily able to evade most objects swatted at them, and Garimella realized that following their flight patterns could help the field of robotics.
Garimella built a small flying robot and developed an algorithm that will allow drones to more efficiently evade objects while flying. His project is one of the Global Finalists for the 2014 Google Science Fair.
The design goal for the robot was to evade moving obstacles, and treat them as threats. The design contraints kept in mind that robots were moving in dynamic situations where the obstacles were complex, unknown and unpredictable.
Mihir decided that the lack of repeatable and efficient threat evasion was holding back true autonomous and scalable flying robots. His first step to address this problem was to design the drone.
Mimicing the fruit fly's physiology he built a sensor module that was lightweight and fast with an accurate system of visual feedback. Two infrared distance sensors send information to the module to understand the location and speed of the approaching threat.
Algorithms were built to follow the fly's behavior, taking into account the escape methods and trajectories used to evade objects.
A great part of the Google Science Fair video is the testing. Garimella shows that he did extensive testing with a plywood sheet to simulate an approaching threat. Each test performed was recorded and video frames were compared to the evasion methods of the flies.
Using biological methods to solve robotics problems is amazing, and the fact that a fourteen year old from Pittsburgh has built this project is even more exciting. The idea of insects inspiring robotics has recently been covered here at ENGINEERING.com but building complex evasion algorithms is another step in the right direction.