In the FEROX project, the technology meets the wilderness in a challenging mission to uncover nature's hidden treasures: various wild berries scattered inside Scandinavian forests. In this tech endeavour, autonomous drones take center stage as they carefully navigate through the woods, armed with AI-powered visual sensors on the lookout for the fruits concealed in the forest’s ground cover. 

Imagine this: drones swiftly gliding beneath the dense canopy, mapping out predefined areas with advanced sensing devices, all without a single human input. Their task? To pinpoint the rich concentrations of fruits that lie hidden in the vast timber. Later on, their findings will assist people in their harvests. The key information in this mission revolves around location: where are the berries? How to reach them quickly and without much effort? Where am I? Lastly, where should I not go? 

Although passing throught the same forest, human and robots use different means to perceive their surroundings. Naturally, their organic and cybernetic brains process this data in their own, particular way, catering to their different needs.  

In this blog post we will unveil the secrets behind assisting both pickers and drones on their quest for fantastic berries. Discover how we bridge the gap between the natural instincts of human foragers and the mathematical precision of our airborne companions, ensuring a seamless collaboration in the search of nature's treasures. Get ready to explore the future of wild berry foraging, where innovation meets the nature. 

Beyond Satellite Navigation 

Before we embark on the journey of navigation, it's crucial to acknowledge the limitations of traditional GNSS (Global Satellite Navigation Systems, which include GPS and other similar technologies), especially when venturing into the heart of the forest. While GNSS may provide a reasonably accurate position information for humans to understand in which part of the forest they are, it falls short when it comes to the precision required by autonomous robots navigating through dense vegetation. The precise location awareness is crucial for the safe and efficient movement of these mechanical companions, especially during the surveying mission, to ensure the complete coverage of the area of interest. 

To overcome the shortcomings of GNSS inside woodlands, drones in the FEROX project employ a sophisticated combination of stereo cameras, Inertial Measurement Units (IMUs), and advanced data fusion techniques. These elements work in harmony to perform Visual Simultaneous Localization and Mapping (V-SLAM1), allowing the drone to create a real-time 3D map of its surroundings while tracking its own location. This approach empowers the drone to autonomously navigate beneath the forest canopy, ensuring it remains on the right track to perform its mission. 

V-SLAM obtains the drone trajectory and a sparse point cloud of the forest
even for complex paths and challenging scenes 

On the human side, moving through the forest is second nature. With our eyes we effortlessly sidestep obstacles like trees, bushes, and holes. However, for our robotic counterparts, such tasks demand a different set of skills. Drones, equipped with perception sensors like stereo cameras or laser scanners, can understand their surroundings in a three-dimensional space and detect potential obstacles. These sensors act as the electronic eyes, enabling the drone to make informed decisions about its route, just as a human instinctively manoeuvres through the forest. 

Now that we know where we are, let’s explore the next step of our journey: where shall we go? 

Drone uses point cloud data acquired from a laser scanner to map the surface (grey) and the obstacles, such as trees 

Mapping the treasures 

Embarking on an aerial voyage through the intricate forest, our drones look out for berries hidden in the bushes. While AI models process the data to detect various berry species in the images, V-SLAM aided by the infrequent appearance of reliable GNSS readings constantly estimates the drone’s trajectory, adding geolocation to the identified fruits. This fusion of technologies sets the stage for an elaborate process of mapping the berry treasures. Geographical coordinates added into the images captured during this woodland odyssey enable creating a spatial database of wild fruits’ locations, which can be represented as density maps. 

At this point pickers can step into the woods, ready to utilize digital maps to aid our harvest. The berry density maps become their guide. The FEROX mobile app does not only mark the locations of berries but also paves the way for navigating them towards these fruitful discoveries. 

One of the FEROX drones flying through the Finnish forest 

Hiking for berries 

Unlike our robotic companions, humans require only approximate guidance to navigate the woodland trails. We naturally avoid obstacles and find the best paths in our close proximity. However, long-range routing can greatly help making the journey easier. Imagine a walk less exhausting, steering clear of steep inclines, and a path that makes the harvest both efficient and swift. 

To achieve this, we harness the power of open-source Digital Elevation Models (DEM), supplementing them with data from OpenStreetMap (OSM) and other open databases. The fusion of DEM and OSM data sets the stage for a least-cost path analysis, allowing us to chart the most energy-efficient route, guiding humans through the forest with the well-informed suggestions needed for an optimal foraging experience. 

Conclusions

Integrating information gathered by drone assists finding wild berries inside a mobile map app