IIT Kanpur, India
Controlling a Robot with Your Mind
May–Jul 2017
5 mins read
What if a computer could read your thoughts directly? That was the original ambition — we set out to decode specific thoughts from brain signals, but the technology (and three months) took us somewhere more achievable: using distinct brain signals to control a robot's movement instead.
The starting goal was ambitious — capturing thought itself. I gathered raw EEG signals from an electrode array on the scalp, filtered the noisy signal using an IIR Butterworth filter, then extracted the features that could actually distinguish one mental task from another: waveform length and autoregression coefficients. Reliably decoding specific thoughts turned out to be a much harder problem than the timeline allowed for.
So the scope shifted to something more concrete and demonstrable: classifying distinct mental tasks using Ensemble Subspace KNN and other ML algorithms, then using those classifications to control a robot's movement in real time. The pivot from "read my mind" to "move this robot with my mind" was itself the real lesson — knowing when to scale ambition down to what the data and timeline can actually support. The project won Best Overall Project of Summer 2017 from IIT Kanpur's Science and Technology Council.
