Going to work in an autonomous drone
CAPE TOWN – Drones are not only being sold in record numbers, but drone technology is developing at a rapid pace. Drone designers and software developers are constantly releasing new possibilities to make drones smarter and more efficient with advanced capabilities.
Even basic drones have magnetometers, gyros, accelerators, GPS modules and processors similarly to those used in our smart phones. But the new smart drone generation goes further with more powerful on-board processors, AI driven software, highly accurate sensors, energy efficient motors, and build-in compliance technology for flight safety. Numerous imaging software advancements combine RGB, infrared thermal imaging, orthomosaic (programmatic matching of individual images to form a new composite image), and data from Internet of Things (IoT) sensors opening up new flight abilities.
The company 3DR recently launched a smart personal drone called Solo, which has an incredibly advanced autopilot system. Two 1 GHz computers and artificial intelligence (AI) algorithms control the drone and allow it to fly without input or control from the remote operator once the destination has been keyed in. The drone is guided by smart sensors and computer vision systems and has object detection and collision avoidance capabilities. This kind of technology is a breakthrough since it will eventually enable the automatic delivery of goods and people.
With all the current discussions around Unmanned Aircraft Traffic Management (UTM), Urban Aerial Transport and flights Beyond Visual Line of Sight (BVLOS), a core component of these advanced, futuristic operations will surely be the improved ability of drones to recognise and react to their surroundings. Until now this was one of the biggest hurdles for drone technology - the development of systems to prevent them from crashing into people, property, or manned aircraft. The sense-and-avoid technology must be sufficiently powerful to reliably detect the presence of objects in real-time, but still energy efficient and small enough to fit into a limited space.
A new tendency is therefore to move away from on-board systems that increase weight and demand more energy by rather streaming video back to an object recognition server. This enables drones to constantly analyse the flight landscape and act according to specified sets of pre-defined instructions.
Complex Pattern Recognition Software acts very much in a similar way, streaming data from the drone to the server and also from the server back to the drone. Fully autonomous Mission Planning Software will enable the operator to change settings; switch to GPS guided mode, and use aerial mapping tools.
The enthusiasm for this new drone technology is so high, that a community of more than 1200 open source developers worldwide are currently re-writing software algorithms that control flight for consumer and commercial drone projects. Formed in the second half of 2014, Dronecode brings together numerous existing open source drone projects and assets under one umbrella organisation governed by the well-known Linux Foundation.
Other developments in the flight control arena is, for instance, the SkySpecs Guardian that is a seamless co-pilot that takes over flight control when an obstruction is close or a collision is impending. Once the risk has been avoided, Guardian fully relinquishes control to the human operator.
DJI, one of the leading companies in the drone industry, developed the basic Matrice 100 quadcopter with a Guidance system, which uses multiple stereo and ultrasonic sensors to automatically avoid obstacles within 20 meters. Guidance can also hover the Matrice 100 with centimetre accuracy above the ground.
At the highly acclaimed Massachusetts Institute of Technology (MIT), researchers created a drone that could fly in very harsh weather conditions by monitoring its own safety. The drone monitors possible damage to its propellers, cameras or sensors and can take an alternative route to minimize potential damage or fly to a place of safety.
Another major challenge in the development of drone technology remains limited power and flight times, although smaller more powerful batteries have been developed to extend the flying time of drones from 30 minutes to 1 or 2 hours. However, fuel cells currently being developed by a company in Singapore, Horizon Unmanned Systems (HUS), may be a solution to power and flight time limitations. They have developed a Hycopter, which is a hydrogen fuel cell powered multi-rotor drone that can stay in the air for 4 hours or 2.5 hours when carrying a 1 Kg payload.
Sunlight Photonics from New Jersey developed a drone, the Sunlink-5, which flies entirely on solar power. But it seems if the most promising solution to the flight time problem currently is hydrogen, which are more efficient than electric of fossil fuel powered drones. Since hydrogen fuel cells rely on air for a reaction and energy, a hydrogen tank generates far more energy than a lithium polymer battery of similar weight.
Recently, a UK research group flew a 20 kg multi-rotor drone powered by a hydrogen fuel cell for over one hour, while South Korean drone company MetaVista used a hydrogen fuel cell to fly a smaller drone for over ten hours. Intelligent Energy, the leader in Hydrogen fuel cells, has made these extended flight times possible by succeeding to manufacture extremely light-weight and efficient fuel cells and very light tanks from modern composite materials to withstand the pressure required.
For us as human beings rest is a normal part of our lives. But drones that rest are somewhat odd. But there have been some interesting developments in this space. A number of researchers have been finding ways to make drones behave more like animals, to see if these changes in design can make them more energy efficient.
An international team of engineers recently published a paper exploring whether it is possible to design drones capable of “perching” or maintaining altitude without consuming energy or deviating from their data-gathering mission. To enable the drone to “perch” it needs to be fitted with an adaptable landing gear that allows it to rest on street lights and the edges or corners of buildings, in order to improve vision stability and maintain the scope of that vision.
The Aerial Robotics Labs from the Imperial College London already in 2017 modified a basic DJI Matrice 100 to shoot out synthetic material like a spider and attach itself to nearby walls, by reeling in the thread and make it taut so it could shut off its motors to save power while “resting”.
Another key component of smart drones is cameras. With new 3D sensing technology and improved gimbal designs, cameras are not only used for aerial photography, but also for orientation, navigation and object recognition.
As with all digital cameras, the advances in aerial cameras have been phenomenal. DJI, the largest drone manufacturer in the world, announced two new cameras for its high-end Inspire drones. The drones are equipped with the first micro four-thirds (MFT) mirrorless cameras designed specifically for aerial use and will offer drone operators 13 stops of dynamic range, a high 16-megapixel resolution and the ability to shoot a 24 or 30 frames per second (fps) 4k quality video. And of course, they are capable of high quality standard still pictures as well.
There are still many obstacles to overcome before smart drones could “uber” us all over the city with great ease and speed. In particular the safety issues will have to be resolved and the avoidance technology needs more refinement. But one thing is very certain: Drones will become more common in the near future. Over the next few years, drone technology will become improved in every way — providing safer, more automated and quieter drones, as they gain capabilities we cannot even imagine today. AI software will make drones much smarter and enable them to analyse and interpret their surroundings and act accordingly.
And perhaps, somewhere in the future when the traffic is even more dreadful than present, you might “uber” to work in an autonomous drone.
Professor Louis Fourie is the deputy vice-chancellor: knowledge & information technology - Cape Peninsula University of Technology.
BUSINESS REPORT ONLINE