Safe Deep Neural Network-Driven Autonomous Vehicles Using Software Safety Cages
Deep learning is a promising class of techniques for controlling an autonomous vehicle. However, functional safety validation is seen as a critical issue for these systems due to the lack of transparency in deep neural networks and the safety-critical nature of autonomous vehicles. The black box nature of deep neural networks limits the effectiveness of traditional verification and validation methods. In this paper, we propose two software safety cages, which aim to limit the control action of the neural network to a safe operational envelope. The safety cages impose limits on the control action during critical scenarios, which if breached, change the control action to a more conservative value. This has the benefit that the behaviour of the safety cages is interpretable, and therefore traditional functional safety validation techniques can be applied. The work here presents a deep neural network trained for longitudinal vehicle control, with safety cages designed to prevent forward collisions. Simulated testing in critical scenarios shows the effectiveness of the safety cages in preventing forward collisions whilst under normal highway driving unnecessary interruptions are eliminated, and the deep learning control policy is able to perform unhindered. Interventions by the safety cages are also used to re-train the network, resulting in a more robust control policy.
KeywordsAutomatic control Autonomous vehicles Cyber-physical systems Deep learning Safety
This work was supported by the UK-EPSRC grant EP/R512217/1 and Jaguar Land Rover.
- 1.Adler, R., Feth, P., Schneider, D.: Safety engineering for autonomous vehicles. In: 2016 46th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Workshop (DSN-W), pp. 200–205. IEEE (2016)Google Scholar
- 2.Archer, J.: Indicators for traffic safety assessment and prediction and their application in micro-simulation modelling: a study of urban and suburban intersections. Ph.D. thesis, KTH (2005)Google Scholar
- 4.Department for Transport: Research on the Impacts of Connected and Autonomous Vehicles (CAVs) on Traffic Flow: Summary Report (2017)Google Scholar
- 7.Heckemann, K., Gesell, M., Pfister, T., Berns, K., Schneider, K., Trapp, M.: Safe automotive software. In: König, A., Dengel, A., Hinkelmann, K., Kise, K., Howlett, R.J., Jain, L.C. (eds.) KES 2011. LNCS (LNAI), vol. 6884, pp. 167–176. Springer, Heidelberg (2011). https://doi.org/10.1007/978-3-642-23866-6_18CrossRefGoogle Scholar
- 8.International Organization for Standardization: ISO 26262: Road vehicles-functional safety. International Standard ISO/FDIS (2011)Google Scholar
- 9.IPG Automotive GmbH: Carmaker: Virtual testing of automobiles and light-duty vehicles (2017). https://ipg-automotive.com/products-services/simulation-software/carmaker/
- 12.Kuffner Jr., J.J., Anderson-Sprecher, P.E.: Virtual safety cages for robotic devices, US Patent 9,522,471, 20 December 2016Google Scholar
- 13.Kuutti, S., Fallah, S., Bowden, R., Barber, P.: Deep Learning for Autonomous Vehicle Control: Algorithms, State-of-the-Art, and Future Prospects. Morgan & Claypool Publishers, San Rafael (2019)Google Scholar
- 15.Lu, J., Dissanayake, S., Xu, L., Williams, K.: Safety evaluation of right-turns followed by u-turns as an alternative to direct left turns: Crash data analysis. Florida Department of Transportation (2001)Google Scholar
- 16.Montanaro, U., et al.: Towards connected autonomous driving: review of use-cases. Veh. Syst. Dyn. 57, 1–36 (2018)Google Scholar
- 17.Polycarpou, M., Zhang, X., Xu, R., Yang, Y., Kwan, C.: A neural network based approach to adaptive fault tolerant flight control. In: Proceedings of the 2004 IEEE International Symposium on Intelligent Control, pp. 61–66. IEEE (2004)Google Scholar
- 18.Ross, S., Gordon, G., Bagnell, D.: A reduction of imitation learning and structured prediction to no-regret online learning. In: Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics, pp. 627–635 (2011)Google Scholar