Journal article
Uncertainty quantification with statistical guarantees in end-to-end autonomous driving control
- Abstract:
- Deep neural network controllers for autonomous driving have recently benefited from significant performance improvements, and have begun deployment in the real world. Prior to their widespread adoption, safety guarantees are needed on the controller behaviour that properly take account of the uncertainty within the model as well as sensor noise. Bayesian neural networks, which assume a prior over the weights, have been shown capable of producing such uncertainty measures, but properties surrounding their safety have not yet been quantified for use in autonomous driving scenarios. In this paper, we develop a framework based on a state-of-theart simulator for evaluating end-to-end Bayesian controllers. In addition to computing pointwise uncertainty measures that can be computed in real time and with statistical guarantees, we also provide a method for estimating the probability that, given a scenario, the controller keeps the car safe within a finite horizon. We experimentally evaluate the quality of uncertainty computation by three Bayesian inference methods in different scenarios and show how the uncertainty measures can be combined and calibrated for use in collision avoidance. Our results suggest that uncertainty estimates can greatly aid decision making in autonomous driving.
- Publication status:
- Published
- Peer review status:
- Peer reviewed
Actions
Access Document
- Files:
-
-
(Preview, Accepted manuscript, 2.0MB, Terms of use)
-
- Publisher copy:
- 10.1109/ICRA40945.2020.9196844
Authors
- Publisher:
- IEEE
- Journal:
- 2020 IEEE International Conference on Robotics and Automation (ICRA) More from this journal
- Pages:
- 7344-7350
- Publication date:
- 2020-09-15
- Acceptance date:
- 2020-01-31
- Event title:
- International Conference on Robotics and Automation (ICRA2020)
- Event series:
- International Conference on Robotics and Automation
- Event location:
- Paris, France
- Event website:
- https://www.icra2020.org/
- Event start date:
- 2020-05-31
- Event end date:
- 2020-06-04
- DOI:
- EISSN:
-
2577-087X
- ISSN:
-
1050-4729
- Language:
-
English
- Pubs id:
-
1079475
- Local pid:
-
pubs:1079475
- Deposit date:
-
2020-03-06
Terms of use
- Copyright holder:
- IEEE
- Copyright date:
- 2020
- Rights statement:
- Copyright 2020 IEEE
- Notes:
-
This is the accepted manuscript version of the article. The final version is available from IEEE at https://doi.org/10.1109/ICRA40945.2020.9196844
If you are the owner of this record, you can report an update to it here: Report update to this record