Using only on-board sensors and cameras, researcher Julián Estévez from the Computational Intelligence Group (GIC) at the University of the Basque Country (UPV/EHU) has developed an affordable, autonomous navigation technology designed to prevent mid-air collisions between drones with intersecting paths. His results have been positive and promising.
A study involving a set of drones has confirmed that, despite the low cost of the technology, the developed solution has been successfully tested on commercial drones. “By utilizing simple, affordable equipment and an algorithm based on artificial vision and color identification, we have created a robust technology that effectively prevents drone collisions and can be applied to most commercial and research aerial robots. We have also made the complete software code for this solution publicly available,” said Estévez.
The findings are published in the journal *Aerospace Science and Technology*.
Most drones we encounter are operated remotely, but for a drone to be fully autonomous, it must make flight decisions independently, without human intervention. This includes avoiding collisions, maintaining its course against wind gusts, controlling flight speed, and navigating around obstacles like buildings and trees.
“This work represents a step toward fully autonomous navigation, allowing drones to independently decide on maneuvers and directions to avoid collisions with each other or with other airborne obstacles. As commercial drone use is expected to increase, our contribution is a small but significant advance in this area,” Estévez explained.
He added, “Our collision prevention approach does not require drones to communicate with each other; instead, they rely solely on their onboard sensors and cameras. We process the camera images to adjust the drones’ reactions, ensuring smooth and precise flight.”
In their experiments, the researchers aimed to replicate realistic drone conditions, including typical urban scenarios with uncontrolled lighting and drones flying in various directions, to ensure their findings are applicable to real-world situations despite the initial laboratory testing.
Color-Based Algorithms
“We outfitted each drone with a red card that helps the software algorithm detect nearby drones and assess their proximity,” Estévez explained. “The concept is straightforward: each drone has an onboard camera that is divided into two sections (left and right). This camera constantly scans for the red color of the cards.”
“By processing the images, we can determine the proportion of the screen covered by the red color and whether the majority of this red area is on the left or right side of the screen. If the red is predominantly on the left, the drone will steer to the right to avoid a collision. Conversely, if the red is mostly on the right, the drone will move to the left. This process applies to all airborne drones.”
“As the amount of red on the screen increases, it indicates that the drones are on a collision course. When a certain threshold is reached, the drone will initiate an avoidance maneuver.”
“This entire process occurs autonomously, without human intervention. It’s a straightforward method to prevent collisions using low-cost sensors and equipment,” said Estévez. “It’s similar to how a person walking down the street would move to the right if they see someone approaching from the left to avoid bumping into each other.”