An autonomous drone demonstrated by Purdue University computer science professor Aniket Bera can follow a person through a forest, manoeuvre around trees and estimate their movements when they disappear from view. Bera, an associate professor at the university and director of its IDEAS Lab, shared a 44-second video of the demonstration on LinkedIn on October 24, 2026. The drone is shown navigating without GPS or a pre-existing map, using onboard computing to make navigation decisions. The demonstration highlights how autonomous systems could operate in environments where satellite signals are unreliable and obstacles make navigation difficult. The technology could eventually help search-and-rescue teams locate missing people, although its effectiveness in real-world rescue operations has not been established.
Who is Aniket Bera?
Aniket Bera is an associate professor in Purdue University’s Department of Computer Science and director of its IDEAS Lab, which researches robotics, computer vision, artificial intelligence and autonomous systems.Bera earned his PhD in computer science from the University of North Carolina at Chapel Hill in 2017. He previously completed a bachelor’s degree in computer science and engineering at Jaypee Institute of Information Technology in India in 2012.His research focuses on areas including robotics, computer vision, motion planning and human-robot interaction, which are relevant to developing autonomous systems that can navigate complex environments. His work includes autonomous navigation, human-robot interaction, motion planning and multi-agent forecasting, all of which are relevant to developing robots that can operate in complex, changing environments.
How the drone navigates through dense forests
The quadcopter’s demonstration shows it following a person between trees while manoeuvring through a wooded environment. Its navigation system combines environmental sensing with algorithms that help it assess obstacles and plan its movements.The demonstration shows the drone navigating without a pre-existing map of the forest. Technical analysis of the video’s telemetry has identified references to navigation and mapping tools, although the precise configuration of the system has not been fully documented publicly.The ability to navigate through unfamiliar terrain is particularly relevant to forests, where trees and branches create obstacles and satellite signals can be unreliable. Autonomous navigation allows a drone to make flight decisions based on sensor information instead of requiring a human operator to direct every movement.
How sensors help autonomous drones navigate
Sensors help autonomous drones estimate their movement and understand their surroundings. LiDAR, short for Light Detection and Ranging, uses laser pulses to measure distances and can generate a three-dimensional representation of nearby objects.Such information can help a drone identify obstacles, including trees and branches, and plan a route around them. However, the precise combination of sensors used in the Purdue demonstration has not been fully established by the available public information.LiDAR-based aerial navigation has been explored in previous research. In a paper published in Science Robotics on January 29, 2025, researchers at the University of Hong Kong described SUPER, a navigation system designed to help aerial robots move through unfamiliar environments while avoiding obstacles.
How the drone tracks people when trees block its view
One of the reported capabilities of the Purdue demonstration is occlusion prediction. Occlusion occurs when an obstacle, such as a tree trunk, temporarily blocks the view of a person.When a person disappears from view, an occlusion-prediction module can estimate their likely movement using previous observations. This may help the drone anticipate where the person could reappear, although a predicted position is not a confirmed location and tracking accuracy can vary with the environment.This capability could be useful in forests, where vegetation frequently interrupts the line of sight between a drone and its target. However, a predicted position is an estimate rather than a confirmed location. Sudden changes in direction or movement can make predictions less accurate.
The importance of onboard computing
The demonstration is presented as operating without cloud connectivity, with navigation computations performed onboard the drone. Local processing allows an autonomous system to analyse sensor information and make flight decisions without depending on a continuous internet connection.This approach is relevant to embodied AI, a field focused on intelligent systems that perceive and interact with the physical world. Such systems combine sensors, computing hardware and software to interpret their surroundings and respond to changing conditions.For autonomous drones, local processing can reduce dependence on communication networks and allow navigation to continue in areas with limited connectivity. It is particularly useful when operating in remote forests or disaster zones where communications infrastructure may be unreliable.
The role of open-source robotics research
The drone’s navigation technology appears to connect with existing research from the University of Hong Kong’s Mechatronics and Robotic Systems Laboratory, known as the MaRS Lab. A technical analysis published by DroneXL on October 26, 2026, identified references to systems including SUPER and FAST-LIO in the video’s telemetry. The analysis also discussed ROG-Map in connection with the mapping system. These tools are associated with research from the University of Hong Kong, although the precise way they are integrated into the Purdue demonstration has not been fully established.A paper on SUPER, published in Science Robotics in January 2025, reported flight speeds exceeding 20 metres per second in experiments with that system. Those results belong to the University of Hong Kong’s research and should not be attributed to the Purdue demonstration. The video illustrates the challenge of applying autonomous navigation to following a moving person through a forest.
How it could help rescue missing people
Autonomous drones with these capabilities could potentially assist search-and-rescue teams looking for missing hikers or people stranded in remote areas.Dense forests can make ground searches difficult because trees obstruct visibility and uneven terrain slows rescuers. A drone capable of mapping its surroundings and navigating between obstacles could help teams search areas that are difficult to reach on foot.Movement prediction could also be useful when a person temporarily disappears behind vegetation. Rather than immediately losing track of the target, the system could estimate the person’s likely direction of movement and attempt to locate them again.The technology may also support disaster-response operations in locations where satellite navigation or communications networks are unavailable. However, its effectiveness in real rescue missions would depend on reliable target tracking, safe flight, battery endurance and performance across different environmental conditions.
Other potential applications
Beyond search-and-rescue operations, autonomous drones could have applications in several other fields.Wildlife monitoring: Drones could help researchers survey remote habitats and monitor wildlife in areas that are difficult to access on foot.Infrastructure inspection: Autonomous aircraft could navigate around bridges, industrial structures and other complex environments while collecting information about their surroundings.Disaster assessment: Following earthquakes, landslides or severe storms, drones could help map damaged areas and identify routes that are difficult for ground teams to access.GPS-denied environments: Robots that estimate their movement using onboard sensors could be useful in dense vegetation, underground spaces and other locations where satellite positioning is unreliable.
Aniket Bera’s research at Purdue University
Bera’s work at Purdue’s IDEAS Lab spans robotics, computer vision, motion planning and human-robot interaction. The forest-tracking demonstration brings several related challenges together: understanding the surrounding environment, navigating around obstacles and estimating a person’s movement when they temporarily disappear from view.
