This article contrives a neural-adaptive constrained controller of the cable towed air-ground recovery system subject to terrain obstacles, unmeasurable cable tensions, trailing vortex, wind gust, and actuator saturation. In air-ground recovery system modeling, the towed vehicle's nominal 6 DOF affine nonlinear dynamics and the cable system's finite links-joints dynamics are formulated. To achieve accurate air-ground recovery under terrain obstacles, an asymmetric barrier Lyapunov function-based flight controller of the towed vehicle is proposed, by transforming the terrain obstacles into time-varying constraints on the vehicle's trajectory. Then, to approximate the towed vehicle's lumped unknown dynamics caused by the unmeasurable cable tensions and airflows, several echo state network (ESN) approximators are established for velocity and attitude subsystems. By using the state approximation errors-based neural weights learning strategy and minimal learning parameter technique, these ESNs possess better transient behaviors and lower online computational burden. Furthermore, the actuator saturation is automatically monitored and released, by incorporating a specially designed auxiliary compensating system into the angular rate control law for compensation. The stability of the closed-loop system is analyzed. Finally, numerical simulations under two air-ground recovery scenarios are performed to demonstrate the performance of the proposed controller.