@inproceedings{TTlta18,
    author = {Maymoonah Toubeh and Pratap Tokekar},
    title = {Risk-Aware Planning by Extracting Uncertainty from Deep Learning-Based Perception},
    booktitle = {Proceedings of the AAAI Fall Symposium on Reasoning and Learning in Real-World Systems for Long-Term Autonomy},
    pages = {96--102},
    year = {2018},
    abstract = {The integration of deep learning models and classical techniques in robotics is constantly creating solutions to problems once thought out of reach. The issues arising in most models that work involve the gap between experimentation and reality, with a need for a quantification of risk in real-world situations. In order to translate advances in robot planning techniques that use deep learning to safety-critical applications, strategies must be developed and applied to assess the risk involved with different models. This work proposes the use of Bayesian approximations of uncertainty from deep learning in a robot planner, showing that this produces more cautious actions in safety-critical scenarios. An example setup involving a deep learning semantic image segmentation, followed by a path planner based on the resulting cost map, is used to provide an empirical analysis of the proposed method.}
}