UC San Diego

In this work we seek to find the underpinnings of neuronal control and movement. In doing so, we hope to gain insight into the function of the brain and also inspiration for the development of advanced robotics and neuronal prosthetics. In motor control, there is strong evidence that many different signals in the brain are combined. We propose that these various signals are different control policies and that these signals are sparse (reflecting the simplicity of neuronal computation). Additionally, neuronal time-integration plays a crucial role in motor control. Hence, we propose that biological motor control involves : 1. Sparse control signals. 2. Time-integration of control signals. 3. Combination of multiple control signals. We used these three simple ideas to outperforms previous models significantly better on human data (with p < 0.001)