Experimental validation of mission energy prediction model for unmanned ground vehicles

Typical unmanned ground vehicles (UGVs) rely on rechargeable batteries for operation, so it is essential to ensure that the batteries will not be exhausted unexpectedly before the completion of the UGV mission. This demands an accurate model to predict the UGV mission energy requirement. For this purpose, two model estimation methods were proposed in the earlier work [1] using a vehicle longitudinal dynamics model. The first method used the least squares estimation (LSE) without using mission prior knowledge. The other method used Bayesian estimation to consider mission prior knowledge (e.g., road grade and rolling resistance). In this paper, we have validated several aspects of the methods via experiments, which includes: (1) evaluation of the measurement sensor capability, (2) examination of relationship between power consumption and vehicle velocity as well as road grade, (3) investigation of the estimation of UGV internal resistance, (4) study of the effect of different road surface conditions on power consumption, and (5) comparison of the performance between the proposed LSE and Bayesian estimation approaches in predicting the mission energy requirement. Therefore, the vehicle dynamic model has been validated. It has also been verified that the Bayesian estimation method is able to predict UGV energy consumption more accurately than the LSE method.