People affected by Cerebral Palsy suffer from physical disabilities that prevent the control over movement, making the walking difficult or even impossible. Treatment for these patients are orthosis to support their legs. While passive orthosis are nowadays quite common, active assistive orthosis are not preferred due to design limitations in the usage and their high cost. ARGO, the Active Reciprocated Gait Orthosis we developed, is a device that overcomes some of the limitations of the state of the art. It is realized from a passive commercial reciprocated gait orthosis, applying sensors and pneumatic artificial muscles to it. With ARGO we aim at reducing patient's fatigue during orthosis usage, and at the same time, collect informations for a clinical validation of the treatment. To move limbs in a non coercive way, following user's intention to move, ARGO deploys an innovative force-based control. It exploits both a neural network and a PID to obtain a fast and oscillations-free responses. In this article we introduce ARGO and present details and results of our high-performance control system.