A group of researchers from Zhejiang University led by Xin Zhou has taken a step towards that reality described by films such as Star Wars , Blade Runner or Prometheus where multi-robot aerial systems are capable of navigating through skyscrapers and spaceships. The test took place in different scenarios, including a dense bamboo forest , where self-driving drones are juggled in flight .
Self-driving drones: problems and challenges
The scholars, in an article published in Science Robotics , highlighted the characteristics and requirements that they had to take into account in order for their project to go through. These are aspects such as obtaining the ideal trajectory, the possibility of extending performance to different specific assigned objectives, all combined with the need to carry out these tasks in absolute (but high) computing economy on board devices as small as possible. .
The proposed solution
After searching through a number of applications, Xin and his collaborators realized that the key to success in achieving the maximum compromise of the aforementioned requirements was trajectory planning . This planning not only allows the modification of the shape of the route but also regulates the time profiles between the various drones.
All this has been achieved thanks to the adoption of a very specific resolution scheme . In fact, through this algorithm, the specific requirements of the task of generating a trajectory can always be formulated as objectives to be achieved: minimization of flight time, maximization of its regularity, obstacle avoidance, reciprocal avoidance between drones , feasibility of flight dynamics. .
For the location of the air swarm, with the exception of the trajectory planning system, a system has been adopted that combines the information coming from the motion detection hardware with the computerized analysis of the scene visible to the camera of the same (VIO, Visual -inertial odometry)
Each drone has been equipped with a whole series of sensors equipped with complete perception, localization, planning and control capabilities and is freely coupled by a transmission network that shares the trajectories.
Intuition from nature and tests
A drone swarm, in general, behaves similar to that of a squadron of insects or a flock of birds . Reasonably, the proposed system is similar to birds able to fly freely through the forest avoiding obstacles and other moving creatures. For example, in short-range navigation, birds rely primarily on the eyes and their vestibular system (hence the improved visual-inertial odometric system, developed by Xin and collaborators).
In addition, the birds adjust the course and speed simultaneously to avoid collisions considering the flight time and fluidity to save energy. This is the reason for the joint optimization of space-time trajectories with multiple objectives.
Add to this the advantage of the electrically powered high fidelity wireless artificial communication system for trajectory sharing and high speed computation for fast planning.
Thanks to all these features , the researchers carried out successful tests in four different situations : group flight through a bamboo forest, flight in a crowded environment, assessment of avoidance with other drones and also the ability of the swarm to follow the guide. of a person if their vision is blurred.
Self-driving drones: uses and future developments
The academics of Zhejiang University have left ample room for maneuver for those wishing to deepen and expand the self-driving drone project . Their hope is that their approach will serve in tasks such as disaster relief, assistance in biological studies, or ready-to-use transport for rovers and drones landing on Mars, and ultimately as collaborative transport use.
Article byAntonio Aversano
The article Swarm of self-driving drones through uncontrolled set ups was written on: Tech CuE | Close-up Engineering .