| Neural engineering is an ever expanding field attempting to restore lost function by injecting signals into or reading signals from the nervous system. Brain machine interfaces (BMIs) are a form of neural prosthetic designed to allow patients to control external devices using signals from cortical neurons. Such systems range from electroencephalogram based systems to those using single neuron recordings.; Despite an explosion in popularity, the neuroprosthetics/brain machine interface field remains tied to experiments and demonstrations much like those conceived at its inception almost 35 years ago. Research in this report attempts to implement a new paradigm in the BMI field, the neural control of an intelligent vehicle platform as a decision process. This interface is unique in a number of ways. First, it implements control of a mobility device as opposed to a robot arm or cursor. Second, it implements a real-time classifier allowing the motor cortex to produce supervisory commands. Third, it intentionally seeks to use the smallest number of neural signals possible without placing any assumptions on the model of modulation used by the underlying cortex. Decisions to pursue these unique aspects stem from assumptions about how a practical clinical BMI system could be implemented. The studies covered in this document attempt to devise, test, and validate an animal model for a new brain machine interface paradigm.; Results of using a system under these assumptions are presented showing five animals in a closed-loop two direction task are capable of completing 71--94% of trials correctly. Further analysis shows that dramatic adaptation in behavior, neural firing patterns, and even neural information takes place while using such an interface despite the fact that the animals are only exposed to the interface for 45 minutes at a time. |
