With increasingly accelerated technological advances, it is becoming unthinkable today to navigate without the help of GPS. It can be done, but by resorting to traditional positioning methods such as rudimentary paper topographic maps and compasses. Even by using the observation of celestial bodies such as the Sun, the Moon, or the stars, or based on the characteristics of the terrain. That is why the recent development by a researcher at the National University of Colombia (UNAL) is so important.
What Edwin Alexander Casallas Moreno, a master’s graduate in Mechanical Engineering from UNAL, has just done is build a system based on cameras, sensors, and artificial vision that would allow drones to maintain orientation and reconstruct the environment in real time when satellite signals fail. During tests, the prototype reduced navigation errors by up to 40 percent compared to traditional methods based solely on GPS.
A drone can enter a tunnel after a collapse, travel through an underground mine, or cross a crop field covered by smoke to detect a fire. In these scenarios it depends exclusively on GPS. It is like trying to walk blindfolded: a small error can completely divert the trajectory. Now, Casallas Moreno’s development allows the drone to “understand” the space surrounding it by using visual references from the environment, reported the UNAL News Agency.
With cameras capable of identifying key points
The principle is similar to the way a person orients themselves in an unfamiliar place: they remember a door, a staircase, or a window to know where they are and where to move next, the outlet added. Instead of depending on satellites, the aircraft uses cameras capable of identifying key points in the environment — such as corners, patterns, reliefs, or objects — and constantly compares them to calculate its movement in real time. In this way, the drone reconstructs a “visual map” as it advances and corrects positioning errors during flight.
The navigation architecture proposed by Casallas Moreno uses cameras, sensors, and artificial vision algorithms to maintain orientation even in environments where satellite signals disappear or lose precision. For the tests, the researcher adapted a quadcopter drone equipped with Intel RealSense D435i cameras, motion sensors, and processing systems capable of interpreting the environment in real time.
The prototype was then evaluated in virtual environments developed on ROS and RViz — platforms used in robotics to simulate autonomous navigation — and later in experimental flights carried out in controlled spaces, where it had to move, detect obstacles, and maintain orientation even without a GPS signal. Behind this development are visual odometry systems and simultaneous localization and mapping (SLAM) algorithms, advanced robotics technologies that allow autonomous machines to locate themselves and move through complex spaces.
One more step in the research stage
During the tests, the drone managed to maintain a stable trajectory even without a GPS signal, with margins of error close to one meter compared to the actual navigation route. In other words, even without satellite assistance, the drone was able to calculate its position with an approximate difference of one meter from its real location.
One of the project’s most important contributions was the development of a “distributed control” scheme: instead of depending on a single central computer, different drone modules — such as cameras, sensors, and flight controllers — process information in a coordinated manner and constantly share data in order to make decisions.
The logic resembles a work team more than a single machine giving orders, meaning that while some components detect obstacles, others calculate trajectories, and still others correct flight stability in real time. Although in Colombia these kinds of developments are still mainly in the research stage, in countries such as China there are already similar applications for automated mining, infrastructure inspection, and remote monitoring of industrial operations.
“In China, autonomous systems are already being used to operate mining machinery and support infrastructure projects. In agriculture, their applications range from monitoring crops and applying fertilizers to detecting fires or analyzing terrain conditions in real time. The expansion of these technologies is no longer a distant possibility but a transformation advancing rapidly,” says researcher Casallas Moreno.
The advancement of autonomous systems such as this one would have future applications in emergency response, environmental monitoring, civil works inspection, and industrial exploration.