This paper introduces the LOPES exoskeleton robot, a newly developed device for interactive gait rehabilitation. The robot combines a freely translatable and 2-D-actuated pelvis segment with a leg exoskeleton containing three actuated rotational joints (two at the hip and one at the knee). The joints are impedance controlled to allow bidirectional mechanical interaction between the robot and the patient. The device supports both "patient-in-charge" and "robot-in-charge" modes, enabling the robot to follow or guide the patient's movements. Electromyography (EMG) measurements on one subject show that free walking in the device closely resembles free treadmill walking, indicating task-specific gait training potential. However, current position measurements are not accurate enough for inverse-dynamical gait analysis, limiting the device's use as a gait measurement tool. The prototype design and performance are detailed, including the choice of degrees-of-freedom (DOFs), actuator specifications, and evaluation methods. The evaluation results demonstrate the feasibility of unhindered walking and the ability to impose gait patterns, with limb orientations agreeing well for control purposes. Preliminary EMG results show minimal deviation between treadmill and LOPES walking, suggesting the device's potential for clinical applications.This paper introduces the LOPES exoskeleton robot, a newly developed device for interactive gait rehabilitation. The robot combines a freely translatable and 2-D-actuated pelvis segment with a leg exoskeleton containing three actuated rotational joints (two at the hip and one at the knee). The joints are impedance controlled to allow bidirectional mechanical interaction between the robot and the patient. The device supports both "patient-in-charge" and "robot-in-charge" modes, enabling the robot to follow or guide the patient's movements. Electromyography (EMG) measurements on one subject show that free walking in the device closely resembles free treadmill walking, indicating task-specific gait training potential. However, current position measurements are not accurate enough for inverse-dynamical gait analysis, limiting the device's use as a gait measurement tool. The prototype design and performance are detailed, including the choice of degrees-of-freedom (DOFs), actuator specifications, and evaluation methods. The evaluation results demonstrate the feasibility of unhindered walking and the ability to impose gait patterns, with limb orientations agreeing well for control purposes. Preliminary EMG results show minimal deviation between treadmill and LOPES walking, suggesting the device's potential for clinical applications.