Get started Bring yourself up to speed with our introductory content.

Autonomous drones come with challenges and great potential

Autonomous drones represent a category of drones where the pilot is no longer actively involved in flying the drone and is instead monitoring it remotely and only taking control when absolutely necessary.

Autonomous drones can fly under GPS-aided navigation. These are autonomous programmed flights with GPS way-points. There are other categories of autonomous drones which can fly in GPS-denied environments or gps intermittent conditions. They can be in the pilot’s line of sight or can be beyond visual line of sight with additional regulations.

Autonomous drones are used across various industries and verticals from delivery of goods, industrial asset inspections, surveillance, construction, indoor warehouse applications and inspecting dangerous areas, such as mines where humans have a safety risk.

The challenges of involved technology

Autonomous operations are complex from a technical standpoint. While AI-controlled drones with vision carry all the hype, underneath there are various sensors, controllers, and sensor fusion algorithms working in tandem to achieve autonomous flight and also serving redundancy.

Sensors are far from perfect. GPS-aided drones can suffer from loss of GPS signals. Localization amd mapping algorithms in GPS-denied or intermittent conditions have limitations based on environment texture, reflectivity, types and the quality of sensors used.

Battery limitations restrict the total duration of the flight and can become a hurdle for automated paths especially for prosumer category drones with limited payload capacity and flight duration. Loss of battery power in the middle of the flight can cause forced landing in rough areas where recovery of the drone is a challenge.

Autonomous drones operating in GPS-denied areas require complex algorithms involving a fusion of sensors such as laser, cameras, inertial measurement unit for localization and mapping, obstacle avoidance, and path planning. Algorithms can fail due to sensor limitations and this can lead to drones getting lost or colliding. Drones can drift due to strong winds, crash against static obstacles, such as trees, and dynamic obstacles, such as birds. Sensors can fail under different environment conditions. To be able to handle all of these factors is an enormous challenge.

Some form of wireless radio beacon signal still remains a necessary requirement to periodically convey where the drone is to the remote monitoring operator. Recovery of a drone that is lost or crashed in especially difficult terrain or a dangerous areas is a complicated issue. Adding redundancy in sensors, services also has an impact on cost and operations. Expensive sensors and sensor-based hardware can provide good operation in GPS-denied areas. But drones are always prone to crashes or failures and more expensive hardware in difficult terrains or conditions has a cost impact.

GPS accuracy can become a problem for accuracy

Deep learning with cameras is used extensively in the newer generation of autonomous drones relying on camera vision for autonomous navigation, but AI remains an empirically tuned black box. The amount of training data necessary to tune the model defeats a number of use cases where it’s difficult to fly the drone with pilot manual control to get access to the data. There have been mind boggling advances in this field and will continue to solve more problems.

For remote monitoring of autonomous drones, camera based video transmission is preferred for first-person view, but also requires high bandwidth wireless transmission and can have larger latencies based on the connectivity chosen. 4G and 5G connectivity is applicable where the cellular signals remain in range.

Overall regulations on drones

The Federal Aviation Administration in the U.S and the equivalent in other regions state that there is a 400 feet altitude flight restriction unless specific permissions are acquired for industrial asset inspection. For drones flying outside line of sight, you must have a spotter that maintains direct line of sight. Flight is restricted within 5 miles of any controlled or uncontrolled airstrip or helipad. There are also restrictions on flying near stadiums, during emergency situations and flying over people not involved with flight operations.

Connected drones

Autonomous drones with secure connectivity to cloud does enable various advantages, such as remote fleet management of multiple drones, flight planning for the entire fleet of drones, pushing software upgrades, deploying new AI models to an already deployed fleet of drones, real-time telemetry, and logging. There is no doubt connected drones especially for autonomous operations are going to be extremely useful. Connectivity can vary from cellular, satellite and other types of wireless.

Industrial use of drones

The market for industrial asset and infrastructure inspections is ripe for use of autonomous drones. Whether it’s inspection of cell phone towers, wind turbines, electrical transmission towers, wind mills, bridges, or oil or gas pipelines, autonomous drones have a huge potential. The agriculture industry is already using GPS-guided drones extensively. Site monitoring including construction is another fast growing area of drone use.

The road ahead

Autonomous drones have come a long way. Autonomous navigation algorithms have matured, available on-board compute power for real time operations has shot up, and sensor prices have come down significantly. The problems are not solved entirely, but autonomous operations are becoming more and more feasible at manageable costs for chosen use cases. The industry must focus on picking specific use cases or problems and developing autonomous solutions for them.

All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.

CIO
Security
Networking
Data Center
Data Management
Close