Regaining of the patient’s ability to walk after stroke is an important goal of rehabilitation programmes. The ultimate goal of gait rehabilitation is to empower patients for overground walking. We have previously developed a prototype of a therapist-controlled mobile platform with compliant pelvis support mechanism that enables balance training during overground walking (device E-go). The aim of this pilot randomized controlled study was to explore the usefulness of the E-go in reducing the number of therapists needed during walking training, and to explore the effectiveness of the E-go on walking abilities in severely affected stroke patients. The study included 19 subacute poststroke patients divided into two groups. The experimental group (nine patients) trained to walk with the E-go and the control group trained within conventional physiotherapy programs for 3 weeks. Outcome measures were walking distance and speed, Fugl–Meyer Assessment, Berg Balance Scale, Functional Ambulation Category and the number of therapists needed during training. At the end of the training both groups significantly improved in walking speed, walking distance, Berg Balance Scale and Fugl–Meyer Assessment (P≤0.001), but there were no between-group differences. The experimental group on average needed a lower number of therapists (P=0.040). These findings highlight the potential of the E-go for overground walking training in severely disabled subacute stroke patients.