An in-depth exploration looking at how the pollination process can be re-created for the Mars environment.

Created: February 1st, 2016


Speculative Proposal / Conceptual Design: 

In conceptualizing a prototype, we considered both the individual plant life cycle and the long-term ecosystem of the Mars inflatable. Because this experiment would last at least 18-months, we wanted the plants to be generative, thriving in an ecosystem that evolves (rather than relying solely on materials sent from Earth). In order to drive towards a self-sustaining habitat, and to start experimenting with the idea of nourishment for future Mars astronauts, we looked into methods for automating plant pollination.

For our initial prototype, we focused on the pollination of tomato plants. Genetically modified tomatoes have successfully been grown in space, and they can be a source of fresh food for eventual settlements. They produce bright yellow flowers, which we can easily identify and pollinate. However, our prototype should be extrapolated to other varieties of buzz-pollinated plants.



Our prototype involves two parts:

We use a webcam and computer vision (Processing) to detect flower locations, so we can redirect the pollinator. When the color of the flower for a specific plant is identified, the software reads the pixels of that color (within a certain RGB threshold) and groups them into flowers. Because we are working with tomato plants, we are identifying yellow against the green leaves and other background noise around the flowers.

Separately, we created a pollinator. When a bee lands, it vibrates its wing muscles, nudging the pollen from the male anthers to the female stigma of the flower, pollinating the flower. This is self-pollination, which doesn’t work for all flowers, but works especially well with tomato plants. The pollinator is a soft sphere containing a vibrating motor and simple circuit that, when a flower is detected, vibrates for a small, set amount of time. This simulates the vibration of a bee’s wings, shaking the pollen loose and pollinating the flower.

The Computer Vision portion identifying flower buds on a prototype paper tomato plant
Screen shot 2016 02 01 at 6.48.57 pm.thumb
The Pollinator, complete with 3.3V battery and fuzzy globe with vibe motor inside.
Slack for ios upload.thumb
The Particle Photon Board future Models could be based off of
Photon new 3e010e9f.thumb


Using computer vision to identify properties of plants is a fairly common way of cataloguing and researching plant growth. Using a camera allows us to identify physical properties of the plants and detect properties of light absorption, water absorption, respiration, diseases, etc. The system developed for this prototype is simple to implement.

In many urban and greenhouse environments, farmers use electric toothbrushes as a form of manual pollination. Our vibration machine was designed to mimic the vibrating and material qualities of the bees that are responsible for tomato pollination on Earth. This system could be tuned to even more accurately mimic the vibrating frequency of bees on earth. It could also include components for communication with the computer visualization software, storing and retrieving pollen dust, and organizing the dust-filled pads to enable further testing and experimentation. 

Furthermore, cross-pollination of plants to produce the most genetically viable option is widely practiced on Earth. This idea lead us to recreate the practice on Mars; while we aren't attempting to produce viable farming methods yet, we do want to use these plants as genetic experiments to determine which plants are hardiest on Mars. Many technologies we learned from were drawn directly from the natural process of pollination, looking at how tomato plants cross-pollinate in their natural environments.



Initially we had identified two processes that could pollinate tomato plants, 1) simulating the vibrations of a bumble bee’s wings, and 2) creating controlled wind breezes using fans to spread pollen throughout an encasement. Following these ideas, we had conversations about collecting plant pollen for selective breeding, and discussions regarding the scope of the test product. We arrived at our final product, which we divided into four parts: fabrication of the motor, software development of the motor, software development of the computer vision, flower fabrication and drawings for proof of concept, and documentation of the finished work. The work was divided by Kyle, Nickie, Ruben, and Siyuan, respectively.

The original scope of the project we defined was a little bit larger than the end goal. The computer vision was going to identify and extract individual flowers and trigger the pollinator unit. We removed this function from the final prototype because the communication boards proved to be difficult to use in the time frame available. 



We learned a bit about plant pollination from this experiment. We also discussed how this module, or a machine similar to it, would be integrated into an autonomous mars-based greenhouse.

Additionally, we learned that the prime geometry for the pollen collector is a sphere, which avoids the complications of mechanized rotation.

The scope of this prototype calls back into question our initial research questions: why are we automating plant pollination in the first place, and what kinds of plants do we want to take to Mars? This level of abstraction allows us to think about our prototype within the context of the entire project.


Open Questions and Challenges: 

In this prototype, we have not physically connected the computer vision component with the motorized pollen collector. However, we imagined that the mechanized arm holding the motor would be capable of moving in three dimensions. Another challenge for this automated pollination system would be reading depth of field. Our prototype recognizes color, but we need additional measures to read the Z-coordinate (stereoscopy or range imaging are options).

We also need to have a pollination process for plants that are not capable of self-pollination.
This could also lead to selective breeding and cross pollination, which would add to the merit of this project as a scientific endeavor.

There are also some remaining questions regarding scaling and improving the technology to make it wireless and more intelligent:
- Are there other methods to achieve the same goal, like robotic arms or drone bees?
- How can the prototype scale to include different types of plants selected for the mission?
- How can we use this technology to experiment with selective breeding?
- How energy-intensive is this technology?


Conceptual drawing for prototype. Pollinators attach to robotic arms, which move in circular paths around hydroponic shelves.


Conceptual drawing for prototype. Pollinators attach to cables and move in three dimensions.

Share this Project


48-528 Responsive Mobile Environments

· 1 member

In Spring 2016, this course was offered in conjunction with 62492 ’Mars Habitat: Building an Atmosphere’ with Christina Ciardullo. Together these two courses explored going to Mars from compleme...more

Focused on

An in-depth exploration looking at how the pollination process can be re-created for the Mars environment.