A seated user watching his avatar walking in Virtual Reality (VR) may have an impression of walking. In this paper, we show that such an impression can be extended to other postures and other locomotion exercises. We present two user studies in which participants wore a VR headset and observed a first-person avatar performing virtual exercises. In the first experiment, the avatar walked and the participants (n=36) tested the simulation in 3 different postures (standing, sitting and Fowler's posture). In the second experiment, other participants (n=18) were sitting and observed the avatar walking, jogging or stepping over virtual obstacles. We evaluated the impression of locomotion by measuring the impression of walking (respectively jogging or stepping) and embodiment in both experiments. The results show that participants had the impression of locomotion in either sitting, standing and Fowler's posture. However, Fowler's posture significantly decreased both the level of embodiment and the impression of locomotion. The sitting posture seems to decrease the sense of agency compared to standing posture. Results also show that the majority of the participants experienced an impression of locomotion during the virtual walking, jogging, and stepping exercises. The embodiment was not influenced by the type of virtual exercise. Overall, our results suggest that an impression of locomotion can be elicited in different users' postures and during different virtual locomotion exercises. They provide valuable insight for numerous VR applications in which the user observes a self-avatar moving, such as video games, gait rehabilitation, training, etc.

We investigate how underfoot vibrotactile feedback can be used to increase the impression of walking and embodiment of static users represented by a first- or third-person avatar. We designed a multi-sensory setup involving avatar displayed on an HMD, and a set of vibrotactile effects displayed at every footstep. In a first study (N = 44), we compared the impression of walking in 3 vibrotactile conditions : 1) with a "constant" vibrotactile rendering reproducing simple contact information, 2) with a more sophisticated "phase-based" vibrotactile rendering the successive contacts of a walking cycle and 3) without vibrotactile feedback. The results show that overall both constant and phase-based rendering significantly improve the impression of walking in first and third-person perspective. Interestingly, the more realistic phase-based rendering seems to increase significantly the impression of walking in the third-person condition, but not in the first-person condition. In a second study (N=28), we evaluated the embodiment towards first- and third-person avatar while receiving no vibrotactile feedback or by receiving vibrotactile feedback. The results show that vibrotactile feedback improves embodiment in both perspectives of the avatar. Taken together, our results support the use of vibrotactile feedback when users observe first- and third-person avatar. They also suggest that constant and phase-based rendering could be used with first-person avatar and support the use of phase-based rendering with third-person avatar. They provide valuable insight for stimulations in any VR applications in which the impression of walking is prominent such as for virtual visits, walking rehabilitation, video games, etc.