 we are developing a system to predict knee and ankle motion using wearable sensors. This system will provide target trajectories for a lower limb prosthesis, which could help patients regain mobility. To do so, we have collected a data set of 23 healthy individuals walking through various environments, including public classrooms, a large atrium, and staircases. We found that adding egocentric vision to the sensor data significantly improved the accuracy of our predictions. Specifically, vision improved the root mean square error, RMSE, by 7.9% and 7% for knee and ankle angles, respectively. Additionally, we observed a significant increase in correlation between the predicted and actual angles, from 0.48 to 1.5%. Finally, we found that the benefits of vision can be further increased with more data. This article was authored by Abhishek Sharma and Eric Rumbokas.