Predictability of Human Motion for Mobile Robot Control

S. Thompson and S. Kagami (Japan)


Human motion, motion prediction


This paper has presented an initial study on the predictabil ity of human motion for application to mobile robot path planning. A constant velocity model of human motion is suf´Čücient for most human movement. Such models must balance the degree of uncertainty within the prediction with the utility the predicted motion has for mobile robot path planning. Here we introduce a method to detect when con stant velocity models of motion are likely to fail by detect ing possible environment constraints on future motion. The presence or sudden removal of environment constraints is used as a predictor of future motion that falls outside the simple constant velocity model. A measure of the pres ence of such constraints is developed using the change in range between human pedestrian and obstacles in the envi ronment. This measure is validated by a comparison with constant velocity model prediction errors.

Important Links:

Go Back