An intelligent control technique has been developed that integrates knowledge-based systems and artificial neural networks in order to emulate behavioral aspects of human skill acquisition. With strategies for leaming and the capability to learn, the Robotic Skill Acquisition Architecture (RSA? utilizes transitions between declarative and reflexive forms of processing to enable system adaptation and optimization. The control scheme attempts to parallel the training of an athlete (the robot) by a coach (the designer), whereby the robot learns through experience how to perfect tasks initially specified in a high-level task language. Knowledge-based systems encode neural network leaming strategies, and skill acquisition is associated with the shif? from a predominantly feedback-oriented, rule-based representation of control to a predominantly feedfonvard, network-based form. This paper describes one aspect of this architecture, in particular how rule-based goaldirected task descriptions are used to obtain initial "rough-cut" system performance. The expressive and flexible nature Humans pass through several levels of competence when motor skills are acquired. As movements become more automatic, cognitive system resources are reallocated, affecting the focusing of attention and the ability to adapt [5-111. Various terms are used to distinguish the forms of memory believed to play a role in skill acquisition, including explicit versus implicit memory [5], declarative versus procedural [6], and declarative versus reflexive m. Each of these distinctions involve going from a predominantly cognitive form of processing to a more automatic one. The approach to robotic skill acquisition described in this paper attempts to capture this feature of biological systems. From a functional standpoint, the technique gives robots the capability to learn, and strategies for learning. Ideally, the design process will parallel the training of an athlete by a coach, whereby the robot learns through experience how to perfect tasb initially shown to it by a trainer through examples and described to it with a high-level task language. This philosophy of intelligent control has taken form in the Robotic Skill Acquisition Architecture (RSA2) [12-141, depicted in Fig. 1. of RSA2 goals is demonstrated through their application to the problem of leaming how to control the longitudinal dynamics of an Within an RSA2 controller, declarative and reflexive forms of airplane during approach and landing. Conclusions are drawn processing are implemented using knowledge-based systems and regarding how a biologically inspired control technique can address artificial neural networks, respectively.Knowledge-based issues related to man-machine interfacing, telerobotics, and adaptive components encode neural network leaming strategies, and skill vehicle control.acquisition is associated with the shift from a predominantly