Nepali Researchers Develop AI-Powered Flight Safety System to Prevent Mountain Crashes in Nepal
साउन ७, २०८३ १४:५९
Kathmandu. Imagine a disaster similar to the devastating floods and landslides of September 2024 cutting off an entire village from the road network. With roads blocked, a drone is deployed to deliver emergency medicines and food. Midway through the mission, strong mountain winds suddenly develop, causing the drone to lose stability, crash into a hillside, and fail to deliver the supplies.
While hypothetical, such a scenario reflects the challenges of operating aircraft in Nepal's rugged geography, where steep mountains, the Himalayas, unpredictable weather, and sudden wind gusts have long posed serious risks not only to drones but also to helicopters and airplanes.
According to data from the Civil Aviation Authority of Nepal's (CAAN) Safety Management Department, 13 of the 24 aircraft accidents recorded between 2013 and 2022 were classified as Controlled Flight Into Terrain (CFIT) accidents.
CFIT refers to an accident in which an aircraft, while still under the complete control of the pilot or flight system, unintentionally collides with terrain, mountains, or other obstacles. In most cases, the pilot is unaware that the aircraft is on a collision course until it is too late.
A team of Nepali researchers has developed a technology that could address one of Nepal's biggest aviation safety challenges. After around one-and-a-half years of research, the team has created a system that can analyze surrounding terrain in real time during flight and autonomously navigate an aircraft to safety if it detects a collision risk.
The research was carried out at the Institute of Engineering, Tribhuvan University, with direct funding from the U.S. Air Force. The project was led by Dr. Sudip Bhattarai as principal investigator, with Prof. Mahesh Chandra Luintel and Prof. Surya Prasad Adhikari serving as co-researchers. The research team also included Kamal Darlami, Ayush Bhattarai, Pratibha Bhandari, Nabin Bhandari, and Nischal Paudel.
The project began in May 2023
"Our primary objective is to completely prevent CFIT accidents in Nepal," Dr. Bhattarai said. "We focused on ensuring aircraft remain safe even when visibility ahead is blocked by clouds or fog."
Before launching the project, Bhattarai's team held discussions with the U.S. Air Force to address what they described as the CFIT problem in Nepal. The discussions centered on developing a system that could prevent aircraft from crashing into mountains when pilots lose visibility due to clouds, fog, or other environmental conditions.
To support the research, the team submitted a proposal to the U.S. Air Force seeking funding to develop the necessary flight controller and collision-avoidance technology.
According to Bhattarai, similar CFIT-related research is also underway in the United States. He noted that researchers there have successfully developed technology enabling aircraft to rapidly climb vertically when a collision is imminent, although the system has yet to be fully implemented.
"CFIT causes billions of dollars in losses every year in the United States," he said. "But in Nepal, it is not only about financial loss. Many people have lost their lives."
With Nepal's mountainous terrain presenting different operational challenges, the team chose to develop an alternative collision-avoidance approach designed specifically for manned aircraft operating in the country.
Real-Time Route Generation
The researchers successfully tested a system called In Situ Path Generation, which allows an aircraft, whether piloted or remotely operated, to autonomously generate and follow a new flight path whenever it approaches a potential collision.
The system gives the aircraft decision-making autonomy during critical situations.
It works by continuously comparing the surrounding terrain observed during flight with a Digital Elevation Model (DEM)—a three-dimensional digital map showing the elevation of mountains, valleys, plains, riverbanks, and other land features.
If the system detects that the aircraft is approaching a mountain or another obstacle, it identifies the collision risk and generates 80 new safe waypoints within three seconds, enabling the aircraft to navigate along a safer route. The avoidance strategy varies depending on the situation.
If flying straight at the current altitude is considered safe, the aircraft maintains its course.
If continuing straight could result in a collision, it attempts to avoid the obstacle by weaving while maintaining the same altitude.
If weaving is insufficient, the aircraft climbs to avoid the terrain.
If climbing alone cannot eliminate the risk, it performs a climbing turn.
When no immediate safe route is available, the aircraft enters loiter mode, circling within a designated area until a safe path is identified.
If none of the available options can provide a safe route or if the aircraft cannot identify a viable path toward its intended destination after rerouting, it automatically returns to its original departure point.
Tested in Mustang
The system was tested using a Skywalker X8 unmanned aerial vehicle (UAV) equipped with a Pixhawk 6C flight controller and PX4 autopilot firmware, which enabled the in-situ path generation capability.
Flight tests were conducted in Lete, Mustang, where researchers deliberately flew the aircraft at distances of less than 100 meters from surrounding mountains. In several tests, the UAV was intentionally brought within 50 to 100 meters of cliffs and steep terrain.
"We deliberately placed the UAV on a trajectory that would have resulted in a collision with the hillside," Dr. Bhattarai said. "The controller then autonomously generated a safe path and redirected the UAV."
Using the developed system, the team successfully operated the UAV over a distance of up to 10 kilometers in approximately 30 minutes.
Potential Commercial Applications
According to Dr. Bhattarai, the technology developed through funding from the U.S. Air Force has the potential to be used in commercial aviation in the future.
The research team has already discussed the project with stakeholders, including companies involved in Nepal's aviation sector and the Civil Aviation Authority of Nepal (CAAN). Those discussions identified several technical and policy challenges that would need to be addressed before the technology could be integrated into aircraft.
However, Bhattarai believes the system could be especially valuable for Nepal's emerging commercial drone industry.
He said Nepal has significant demand for drones in sectors such as delivery services, remote sensing, disaster response, search and rescue, and other commercial applications, but the country's challenging geography continues to limit operations.
According to him, long-distance manual drone flights are impractical because mountain valleys and rapidly changing weather make it difficult for pilots to control aircraft from afar.
The challenges become even greater during Beyond Visual Line of Sight (BVLOS) operations, where drones fly beyond the pilot's direct field of view. In such cases, he said, autonomous flight technologies and well-defined Standard Operating Procedures (SOPs) are essential.
"If you try to fly a drone from Pokhara to Mustang, the chances of it completing the journey without an accident are very low because of the mountains along the route and sudden wind gusts," Bhattarai said.
He added that drones operating in such environments need sufficient autonomy to protect themselves whenever they encounter situations that could lead to an accident.
According to Bhattarai, the newly developed system currently operates between autonomy levels 3 and 4.
"This technology enables the aircraft to identify obstacles along its flight path and independently determine a safe route during dangerous situations," he said. "If the industry adopts it, Nepal could see significant growth in the commercial use of drones."
पछिल्लो अध्यावधिक: साउन ७, २०८३ १४:५९
