Expanding Horizons: Elevating Our Machine Learning and Autonomy Development with a New Flight Lab
Last month, we shared more detail on the expansion of our Flight Lab with a modified Beechcraft King Air C90 featuring enhanced avionics and advanced data collection capabilities. This new addition accelerates the testing and iteration of autonomous flight and machine learning (ML) solutions while supporting Acubed’s agile, data-driven product development and autonomy breakthroughs. This raises the question: how exactly is the team approaching the aircraft upgrades and how are they essential for advancing our ML solutions to the next level of engineering readiness?
The King Air is an ideal platform for advancing Acubed’s efforts, offering improved dimensions for mounting cameras and sensors with views similar to an A320. This enhances data collection, enabling the team to refine ML solutions for functions such as autonomous taxiing. Building on the success of the existing Flight Lab, our modified Beechcraft Baron 58, the King Air’s larger size, increased payload, and power allow for higher-quality data and easier testing of new sensors and equipment. Its versatility enables engineers to conduct rapid testing and iteration, including modifying and hot-swapping components as well as reviewing experimental data in-flight.
Avionics Overhaul
The King Air has already undergone a complete avionics overhaul. Among the upgrades is a new instrument panel with a set of modern, digital flight instrumentation, and displays. The digital instrumentation will be integrated with the King Air system, allowing for future development of a closed-loop flight test system.
Nose to Tail
Custom engineered nose and tail mounts will accommodate cameras and sensors used to collect data. A precursor to this system has been in place on the Baron and it has helped gather massive amounts of airport and runway imagery and data from hundreds of locations across the United States. This approach has been used to inform ML algorithms that enable a variety of autonomous functions, leading to breakthroughs in autonomous flight.
The custom nose mount design, for example, allows for modular camera and navigation sensors to be installed in the nose of the aircraft. Ten cameras collect synchronized images in multiple orientations, across various stages of flight, to feed into our data collection platform. The mount platforms also include wide-angle lenses for the outer sensors on both the nose and tail sections of the aircraft, allowing data to be collected for the development of an obstacle detect-and-avoid system in the next phase of our autonomy efforts.
In designing the tail mount, engineers accounted for the existing tail structure, aerodynamic loading, thermodynamic effects, RF signal attenuation, and more, to limit any side effects of the modification. Additionally, the materials used in fabricating the sensor mounts make them resistant to the harsh environmental conditions the King Air will encounter in its data collection campaigns. Seen below is a sample configuration of the vertical stabilizer camera mount. The key benefit of mounting cameras and sensors on the tail of the King Air is their position 14 feet off the ground – roughly the same height as the belly of the A320 – so the images and data mimic the vantage point of a commercial airliner.
Vision for the Future
The Flight Lab is a key component of Acubed’s data-driven product development. From leveraging quantum technology to explore alternative navigational solutions, to collecting data at scale to train machine vision systems, the King Air furthers our mission of developing autonomy solutions for the future of flight.
If you are interested in joining our team, check out our job postings here.