The goal is to have robot locomotion to be controlled by a houseplant. It is possible to record electrical potentials in plants so maybe the signals can be used to control prostheses in a way similar to EEG controlled projects.
The project is heavily based upon others work. The frame is based upon "Stubby the (Teaching) Hexapod" by The Big One, here on Hackaday. The code is written in python by Mithi Sevilla and adapted for the Stubby frame. The code is based upon her code for "Hexy the Hackaday Hexapod". These forked projects are listed in external links.
First draft of the plant sensor. Hopefully it will be sent to pcb printing tonight. It still needs a current regulator and a dac. The circuit design is done by Dewald De Bruyn and is based on the Plant Spikerbox.
Thomas Paniagua have done some clustering with the 1000 data points on my gist. In summary what he did right now is extracted 3000 useful features from the data I gave him and used a visualization algorithm to figure out whether there were meaningful clusters in the data. According to him it turns out there are... and they are very clear. In the attached image you can see a few clusters after he have processed the data, where each region of clusters can be assigned to a plant action.
At the moment a researcher at North Carolina State University is looking into how to interpret the sensor data from the houseplant. His current research deals with this kind of time series data, specifically a Convolutional Autoencoder for working with unlabeled human physiological data which he believes can be applied to plants.
I have also engaged an electrical engineer to improve the data acquisition. I will publish the schematic for this as soon as it is done.
Please make this happen it is actually insanely cool good job!