Pollination within closed system
by Magics Instruments
Robotic nanodrone pollinator
The consortium is preparing for future research on a wide variety of crops to include in baked goods on Mars. A lot of relevant crops rely on pollination by insects. As insects are not present on Mars, Magics Instruments will develop a robotic nanodrone pollinator that is able to recognize flowers and facilitate pollination of future space crops.
Key objectives
- Energy-efficient AI in battery-powered edge devices.
- Object detection using deep learning.
- Path planning and drone control.
- Low-power AI processor design.
Approach
-
Task 1 - Energy-efficient AI model development
In this task, we will develop a light-weight CNN AI model for the detection of the flowers of some specific plant species and the estimation of these flower’s 3D angle in the world. We will change the architecture of the state-of-the-art object detection neural network YOLOv3 such that it can output, next to the center pixel location of each flower in the image, an estimate of the 3D angle direction of the stem of the flower, as illustrated above.
-
Task 2 - Autonomous drone navigation
In the next step, we will use the previously developed flower detection neural network of task 1, and combine information of multiple images to create a sparse flower map of the environment. Using that map, the drone can sequentially approach each of the flowers. After coming as close as possible, the drone backs up again and starts approaching the next flower.
-
Task 3 - Energy-efficient AI processor design
A customized AI chip for flower detection and autonomous drone flight control is needed to deliver the required on-board computational power and energy efficiency. The chip will be implemented in a deep-sub-micron CMOS process, and consists of a neural-network accelerator, a RISC processor, path-planning and motion-control accelerators, and memory blocks.
Meet the Magics Instruments expert
-
Jens Verbeeck
CEO
Jens Verbeeck was born in Turnhout, Belgium in 1984. He graduated as Master in Industrial Sciences, option electronics in 2007 and gained his first professional experience at Intersoft Electronics as electronic hardware engineer. In October 2008 he pursued a Ph.D. degree in Electrical Engineering. His main research topics included the application of electronics in harsh environments. In 2015, he co-founded together with Ying Cao MAGICS instruments NV. MAGICS instruments is a semiconductor company specialized in enabling technologies for intelligent machines and robotics. In his role as CEO he attracted multiple business opportunities and customers, has set up highly technical team and an industrial network capturing multiple opportunities to apply electronics in intelligent machines and robotics.