In the future use of Brain Machine Interfaces (BMIs) it will be important to bring to the subject the feeling of touch when the robotic device grabs an object in the external world under the brain control of the subject. We are now developing the modelling and signal processing infrastructure to implement a somatosensory BMI by stimulating the Thalamus and producing in the primary sensorymotor cortex a response that mimics the normal stimulation felt by the subject. This talk will present the adaptive inverse control scheme and the spike kernel models that we are developing and some preliminary results.