System and method for full-stack verification of autonomous agents

作者: Ros Sanchez German

DOI:

关键词: Autonomous agentStack (abstract data type)Sensory inputNoise (video)Object detectionInformation dataGround truthComputer scienceArtificial neural networkReal-time computing

摘要: A method for full-stack verification of autonomous agents includes training a neural network to learn noise model associated with an object detection module agent system vehicle. The also replacing the and sensory input ground truth information apply surrogate function information. further verifying including trained in response simulate sensor data at least planner system. controlling behavior vehicle using verified module.

参考文章(10)
Javier Alonso-Mora, Scott Frazier Watson, Paul A. Beardsley, Shared control of semi-autonomous vehicles including collision avoidance in multi-agent scenarios ,(2014)
Silvio Savarese, Sebastian Thrun, Jesse Levinson, David Held, Robust Anytime Tracking Combining 3D Shape, Color, and Motion with Annealed Dynamic Histograms ,(2015)
Ossama Emam, Mohamed A. Bahgat, Sara A. Noeman, Ayman S. Hanafy, Vehicle accident avoidance system ,(2015)
Turcot Panu James, Mahmoud Abdelrahman N, Mavadati Seyedmohammad, El Kaliouby Rana, Cognitive state evaluation for vehicle navigation ,(2018)
Jinjun Wang, Zhaowen Wang, Jing Xiao, Kai-Hsiang Lin, Thomas Huang, Substructure and Boundary Modeling for Continuous Action Recognition arXiv: Computer Vision and Pattern Recognition. ,(2012)
Lakshmanan Barath, Guruva Reddiar Palanivel, DEEP LEARNING SOLUTIONS FOR SAFE, LEGAL, AND/OR EFFICIENT AUTONOMOUS DRIVING ,(2019)
Clement Farabet, Gary Hicok, Michael Cox, Greg Heinrich, Curtis Beeson, John Zedlewski, Claire Delaunay, Tony Tamasi, Matthew Campbell, Zachary Taylor, Mark Daly, David Auld, Rev Lebaredian, Training, testing, and verifying autonomous machines using simulated environments ,(2019)