Georgia Tech System Cuts Drone Trajectory Error by 50 Percent

Researchers at the Georgia Institute of Technology have developed a system that allows autonomous aircraft to transcribe and interpret pilot-to-pilot radio transmissions, reducing trajectory prediction errors by more than half. The framework aims to integrate Drones safely into airspace around non-towered airports without requiring human pilots to alter their standard communication procedures.

According to a press release issued by the university on June 23, 2026, nine out of 10 airfields globally operate without active air traffic control towers. At these facilities, pilots rely on shared radio frequencies to announce their positions and intentions. The Georgia Tech team, led by robotics Ph.D. student Sundhar Vinodh Sangeetha and Assistant Professor Sarah Li, presented their findings at the IEEE International Conference on Robotics and Automation in June 2026. Their research addresses a critical gap in uncrewed aerial systems integration by translating natural language radio calls into actionable navigational data.

Translating radio calls into trajectory data

The multimodal framework utilizes speech-to-text software combined with a modified large language model to process standard aviation radio transmissions. By inferring pilot intent from these calls, the system feeds the resulting data into existing trajectory prediction algorithms used by autonomous aircraft.

The researchers validated the system using flight data and recorded radio calls from a non-towered airport in Pennsylvania. The integration of voice data reduced the average error in predicting an aircraft’s flight path from nearly one kilometer to approximately 400 meters, representing a reduction of greater than 50 percent.

Sangeetha stated in the release that human pilots should not have to change their behavior to accommodate autonomous operations. “This is how humans have operated safely for decades around these airports,” Sangeetha said. “So, if we start flying robots here, they should operate in the same way.”

Regulatory context and future applications

The development aligns with ongoing regulatory efforts by the Federal Aviation Administration (FAA) to manage mixed-use airspace. On February 4, 2026, the U.S. Government Accountability Office published a report (GAO-26-107648) highlighting the necessity for the FAA to establish protocols for how drones will communicate with and avoid crewed aircraft within the National Airspace System.

The Georgia Tech team envisions the technology not only as a collision avoidance tool for drones but potentially as an automated safety monitor for non-towered Airports, capable of warning human pilots of impending conflicts before accidents occur.

The researchers are also exploring bidirectional communication capabilities for autonomous systems.

“We’ve shown we can turn language into position. Can we go backwards and generate that language so the autonomous aircraft can announce its intention and coordinate with humans on the same channels?” Li said in the university statement.

AirPro News analysis

We view the integration of natural language processing into uncrewed aerial systems as a necessary bridge for near-term airspace integration. While the FAA and industry stakeholders have heavily invested in digital datalinks and automatic dependent surveillance-broadcast (ADS-B) technologies, the reality of general aviation is that voice communication remains the primary method of deconfliction at non-towered airports. A system that allows autonomous aircraft to passively monitor and interpret standard common traffic advisory frequency calls mitigates the immediate need for costly avionics upgrades across the legacy general aviation fleet. The challenge moving forward will be certifying large language models for safety-critical aviation applications, a hurdle regulators are only just beginning to evaluate.

Sources: Georgia Institute of Technology