Collision avoidance in dynamic environments
Despite their small brains and limited visual systems, flying insects use vision to efficiently detect and avoid collisions with obstacles on a regular basis. These include stationary objects that appear in their flight path, such as trees and bushes, as well as fast-moving objects, such as foraging conspecifics and branches blowing in the breeze. The problem of avoiding obstacles in a dynamic environment is not limited to flying animals however, fish also face similar challenges when controlling their motion in water. In this project we are interested in understanding how both flying and swimming animals use visual information to detect and avoid static and moving obstacles in their environment.
We use behavioural analyses to investigate how animals such as bumblebees and zebrafish use visual cues for maintaining a safe distance from nearby surfaces, avoiding collisions with conspecifics and with other moving obstacles. We present the animals with different forms of static and moving two- and three-dimensional obstacles and record their response using high speed synchronised cameras that allow us to reconstruct the trajectories in three dimensions. The data that we obtain from these experiments are used to develop biological principles for visual flight and swimming control and obstacle avoidance that are then developed into mathematical models of visual guidance that can then be tested in different simulated environments. Our goal is to use these models to inspire the development of lightweight and computationally efficient guidance control systems for robots that are capable of moving safely through dynamic environments.
A comparison of the trajectories of bumblebees and zebrafish moving along an experimental tunnel when we minimised the visual information on one wall. Bumblebees fly further away from the wall that provides strong visual feedback (vertical stripes) but fish react in the opposite way. Understanding the different mechanisms underlying these responses is one of the goals of this project.
- Professor Dario Floreano, The Laboratory of Intelligent Systems, The Swiss Federal Institute of Technology Lausanne
- Dr. Birger Johansson, Cognitive Robotics, The Cognitive Science Department, Lund University
- Julien Lecoeur, The Laboratory of Intelligent Systems, The Swiss Federal Institute of Technology Lausanne