Improving functional motor deficits in patients with neurological and musculoskeletal disorders has been a primary objective of rehabilitation. Task-oriented training in physical environments is a leading approach for rehab training utilizing tasks that are challenging, adaptable, and meaningful. Although different training paradigms can be utilized in a laboratory setting, they often require expensive training equipment and personnel, limiting the number of training sessions. Creating an effective virtual reality (VR) training environment can greatly increase the repetition, cognitive engagement, and training variability. Specifically, dynamic trunk control is required for the successful completion of everyday tasks. These tasks require coordination between the head, upper and lower trunk, and the pelvis. We have developed a novel robotic device, trunk support trainer (TruST) and a VR gaming environment, which allows training in a seated posture at and beyond an individual's point of stability failure using an assist-as-needed force tunnel strategy. The VR environment consists of challenging reaching tasks where the subject's dominant hand is used to manipulate a VR drone. We conducted a study that involved three groups: ten adult subjects trained using VR, ten adult subjects trained in a physical environment (PR), and ten without TruST assistance (control group) to investigate the changes in trunk kinematics. Our study supports the hypothesis that a single training session with TruST in both VR and PR can increase lower trunk range of motion. This approach may be useful for creating training variability and translating therapy in home settings.