SYBIL
















An interactive robotic
soil-spraying installation 
Contact Ars Electronica





















Architecture and Human Augmentation (ETH)
+ MAEID
























SYBIL

Humans and machines increasingly imagine, build, and create together. Yet communication between them often remains one-directional. Humans give instructions, and machines execute them, with little space for interpretation, response, or change. Sybil, an interactive robotic installation at Ars Electronica 2026, explores what a more reciprocal exchange between humans and machines could look like, and what this might mean for the future of design and construction.

Sybil follows a human gesture as it moves through different layers of technological translation. The title refers to the ancient sibyls, oracles who interpreted human questions and returned them in transformed form. Visitors are invited to draw traces into a sand floor using sticks provided at the installation. Cameras and sensors capture these gestures and translate them into robotic toolpaths. A robotic arm then uses these paths to spray a mixture of soil and seeds onto a suspended net structure. At the same time, a feedback system returns aspects of the gesture through sound, image, and data, making the translation process perceptible. Over the five days of the Ars Electronica festival, the sprayed traces begin to germinate and gradually transform the net into a living surface. 

Sybil asks what happens to meaning through each step of this process. A mark in sand is not the same as a robotic trajectory, and a robotic trajectory is not the same as the biological growth that follows. In relation to the festival theme Negotiating Humanity – Future Begins, Sybil understands the future not as something that can simply be predicted, but as something that emerges through the exchange between human intention, computational translation, robotic action, and biological growth.

Concept rendering  of Sybil showcasing the interaction space and growing surface
A robot sprays soil , seeds, and water onto a net structure following flexible input paths.
How do we detect your drawings? We use three depth cameras to read the surface as a depth image. We calibrate the three camera feeds into a shared coordinate system and stitch them into a single top-down view. A groove drawn into the sand appears as a deviation from the flat surface. Once a new drawing is detected, the groove is extracted as an ordered sequence of 3D frames containing position and orientation along the line. Each drawing is then stored with its points and a timestamp.

How do we decide between soil spraying and watering? 
Computationally, the surface is divided into 17 tiles. Each visitor’s drawing is translated and rotated to fit one of these tiles. A managing script (the Manager) runs in the background and decides which tile to use next. The Manager tracks which tiles have recently been sprayed, which are still empty, and which require watering. Soil spraying happens during interaction periods, while watering follows a separate maintenance cycle linked to the time of day and the material's condition.

What happens during the interaction and maintenance cycle?
During the interaction period, the system continuously watches for new drawings. When a visitor creates a drawing and the robot is available, the system assigns the drawing to a tile that has not been sprayed recently. The line is converted into a robotic toolpath that accounts for the robot's reachability and is then executed automatically. The robot sprays the soil mixture along the path the visitor drew. If the system detects no new drawing for some time, it can replay an earlier path instead. During maintenance periods, visitor detection is paused, and the robot works through stored paths autonomously. Soil spraying and watering are coordinated independently, allowing the system to maintain the material and vegetation while preparing the surface for the next interaction cycle.

What are we spraying? The robot sprays a living material mixture. It consists mainly of water, combined with clay, soil, psyllium husk, cellulose, and seeds. The plants growing on the surface include grass, chia, alfalfa, and radish. 

A Sybil is an oracle said to predict the future.  At first, there seemed to be only one Sybil. Three centuries ago, there were at least ten.  

Our Sybil is a robot that interprets visitors' traces and responds by translating them into robotic spray paths. Her prediction is not spoken, but materialized as growth on a net structure, slowly transforming into what might come next.

Fixing the net structure to the steel subtructure.
The net is first secured with cable ties, then woven with acrylic string.
The entire structure is hung from a beam of the Ars Electronica. It does not touch the ground.
On top of the acrylic net is a thinner mosquito net to support the seedlings as they grow and provide more grip for the sprayed paths.
The mosquito net is secured with metal staples onto the acrylic net. 
Last check if the net structure is securely placed before the first layer of soil is sprayed.
The two nets form the soil's substructure. No additional layer is needed as we designed our material mix to stick to thin surfaces.
The first step is a thin layer of soil manually applied. This support the growth. This step was done manually at ars because our robot was still packed but can be done as well robotically.
After that we are placing another layer of thicker soil as a substrate for first growth.
As the seeds are directly in the soil we spray there is no additional need for placing plants. Now we just need patience and tim

Credits


Project by Architecture and Human Augmentation (AHA), ETH Zurich
Prof. Dr. Daniela Mitterberger
Project management Shravan Arun, Fabio Scotto Design and Fabrication: Shravan  Arun
 Living material design Ruth Sidabutar, Tiziano Derme (MAEID)
 Robotic soil spraying Ruth Sidabutar
 Robotic setupCharlotte Thorn, Ruth Sidabutar, Fabio Scotto, Sam Bourgault AR fabricationShravan Arun,  Eleni Alexi UX DesignCharlotte Thorn Interactive SetupDaniela Mitterberger, Charlotte Thorn, Shravan Arun, Linfeng Zhou, Sam Bourgault Expert consultations LAYER3D (Spraying setup)
Prof. Dr. Pierluigi D'Acunto (Structural Engineering)
Carl Stahl GmbHThanks toTeam of Ars Electronica (especially Christl Baur, Veronica  Liebl, Yoryos  Styl, Didi, Tatiana, Thomas and others)
Robotic Fabrication Lab (RFL) ETH Zurich

 Sponsors Department of Architecture - ETH Zurich
Carl Stahl GmBH
MAS DFAB ETH Zurich

All work and images by Architecture and Human Augmentation (AHA), ETH Zurich