Adding intelligence to uncrewed aerial vehicle (UAV) design to improve performance in a variety of dynamic environments offers benefits to many civilian and military mission objectives. Urban areas with tall buildings and vast street systems create drastic changes in the flight environment that are challenging for autonomous flight vehicles to overcome. Forrest fires with large temperature gradients also create difficult wind conditions for a drone to accurately survey the area. Autonomously reacting to these environmental changes would improve the adaptability of a single UAV design, allowing performance in a broader range of conditions and effectively increasing the mission scope for these vehicles. Thus far, the fields dedicated to developing adaptive and intelligent systems for aircraft have remained split. One popular area of focus uses multifunctional smart materials to create unique shape changes in an aircraft structure, known as morphing. Another area of research that is rapidly gaining interest is the field of artificial intelligence and machine learning for controller development. In this work, I brought these two fields together to create an intelligent multifunctional morphing system for autonomous adaptive flight. First, I developed a reinforcement learning (RL) training format for autonomous policy development in a physical hardware environment. The pseudo-episodic training scheme alternates traditional training episodes with exploration episodes designed to randomly reset the following training episode. This format creates space for additional policy updates using off-policy actor-critic and experience replay between traditional training episodes. I tested the pseudo-episodic training format on an airsled-airtrack experimental environment used as a one-dimensional analogy for an aircraft at equilibrium. The inclusion of these additional updates improved learning speed and consistency. Next, I used deep reinforcement learning (DRL) to create two controllers for a macro fiber composite (MFC) actuated camber morphing airfoil. I tested the two DRL controllers on the physical morphing airfoil over a series of step responses when using true state observations from an external sensor, and imperfect state observations from the two state inference models. I compared the performance of the two learned controllers to a traditional proportional-derivative (PD) controller. I found that the learned controllers were faster and more accurate than the PD method and were able to account for the hysteretic behavior inherent to the piezoelectric MFC actuators when provided imperfect feedback. Finally, I used DRL to train gust load alleviation (GLA) controllers for an MFC morphing wing that consisted of three morphing sections. The trained controller used pressure sensors for state observation and learned to reduce the change in lift experienced during upward and downward gusts. I found that increasing from one pressure tap to three pressure taps significantly improved overall controller performance, but increasing from three pressure taps to six pressure taps did not show significant improvement. Overall, the controllers were able to reduce the change in lift experienced by the wing during a gust by 71% to 87%. In summary, the work presented in this dissertation shows a gradual increase in environmental complexity from chapter to chapter in order to develop intelligent controllers for adaptive morphing UAVs. I began with a simple one-dimensional analogy for trim, and ended with an autonomous gust alleviation system trained entirely on hardware using DRL. This work offers an exciting combination of multifuncitonal material morphing with learned autonomous control as a step towards developing truly intelligent morphing UAVs.