Detect & Avoid helicopter (sample image)
Looking at the constantly growing drone market with more and more new manufacturers, fields of application and possible uses, the question of the safety of the soon countless flights quickly arises. Especially during missions out of sight or over populated areas, there are many sources of danger that can lead to accidents. New rules, regulations and risk assessments (SORA) are intended to increase safety here. So are new business models, such as the use of 5G radio networks. But is this really sufficient? Are there enough well-trained pilots for difficult missions? What happens when a radio link or the GPS signal is lost? Questions to which there has been no precise answer so far. Autopilots alone cannot solve the problem, as they are largely dependent on the previously mentioned data streams. Unlike in airplanes or helicopters, pilots do not sit on board to be able to intervene quickly in an emergency. It will also not be possible in the future for air cabs, as a pilot would call into question the economic viability of the entire business model.
MACHINE LEARNING AND SENSOR FUSION AS THE BASIS FOR SECURE AUTOMATION FUNCTIONS FROM SPLEENLAB
It is precisely in this niche that we at Spleenlab are addressing with our new products and developments. Because many of the mentioned questions for safe automated and later autonomous flight can only be solved by efficiently trained algorithms. Artificial intelligence or better machine learning are the most important keywords here, along with sensor fusion.Automated detection of various airspace participants over different distances for safe collision avoidance
Because in order to really safely avoid a collision with airspace participants or suddenly appearing obstacles, clear strategies and automated decisions are needed in real time on board the aircraft. This can only be made possible from an interaction of various mutually safeguarding sensors with a specifically trained AI. For this reason, we have developed two groundbreaking and, above all, operational products with our VISIONAIRY Detect&Avoid and VISIONAIRY Safe Dropzone&Landing. This enables the unmanned aerial vehicle (be it a drone or, later, a pilotless flying car) to react to obstacles even without a GPS signal or after a radio link has been broken, and to simultaneously determine a safe landing zone.Automated and safe landing and drop zone detection (green)
This is made possible, among other things, by a 3D environment map generated in real time. This is used not only for self-localization, but also for so-called semantic environment perception. This enables the aircraft to recognize the actual obstacle and decide autonomously whether to take evasive action or head for a safe landing zone. This is done by means of a precise ground risk analysis, i.e. the detection of particular hazardous situations, such as passers-by in the landing zone. The same applies to the determination of safe drop zones for delivery drones.
DEPLOYABLE PRODUCTS AS A PLATFORM-INDEPENDENT SYSTEM SOLUTION FOR ANY AIRCRAFT
Pilots who can only rely on one camera are prone to errors, which, as mentioned above, can lead to serious accidents in the worst case. Spleenlab is responding to these hurdles with the development of AI-based, certifiable products, for the successful market development of drones and air cabs. Critical to our success is that these products are platform-independent and thus can be integrated on and into any aircraft. They also serve as redundancy to the FLARM collision warning system or the ADS-B airborne surveillance system. This guarantees an additional level of safety.VISIONAIRY CUBES (Sensor Prototypes) by Spleenlab
If an aircraft lacks the appropriate sensor technology, it can be equipped or retrofitted with the functions developed by us thanks to our specially developed VISIONAIRY CUBE – a lightweight, compact plug n‘ play AI sensor suite. Thus, in the future, for the first time, flights out of sight, even under difficult conditions, will be possible with an AI-based safety concept. Initially, this system is still intended to support a pilot or ground control station. However, these safe machine learning applications are the basis for the autonomous flying of manned and unmanned aircraft of tomorrow.