Adaptive Control in Neurorobotics
Automatic control aims at developing algorithms that tend to cancel the error between the goal (input) and the result (output measure), and at optimizing accuracy and response time.
In neurorobotics, neither the input (volitional commands) nor the results (movements) are parameters that can be numerically measured. Instead, they constitute commands that are numerically decontextualized from the environment, as well as measures that not always are geometric, as for instance muscular physiological response. In this field, rather than optimizing accuracy, the goals are to ensure the correct interpretation of the orders given by the user and to achieve maximum insensitivity in front of disturbances, as tremor or spasms. Response time in neurorobotics is not a critical issue. In addition, these control systems should be endowed with environment perception capabilities in order to achieve a high degree of user’s safety.
After a brief description of the main issues involved in a rehabilitation process, its main characteristics and requirements, this lecture will mainly focus on the problematic derived from the high heterogeneity of the actions to be performed and to the complexity of the environment, as well as on the hierarchy of the orders given by the user. Some hybrid control strategies will be described, related to closed loop systems which parameters, affected by noise and delays, are either numerically expressed or correspond to physiological signals. An overview of the state of the art in this application field will complete the exposition, especially remarking the progress in neurological signals based volitional control.
