Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Development and testing of an image transformer for explainable autonomous driving systems

Development and testing of an image transformer for explainable autonomous driving systems Perception has been identified as the main cause underlying most autonomous vehicle related accidents. As the key technology in perception, deep learning (DL) based computer vision models are generally considered to be black boxes due to poor interpretability. These have exacerbated user distrust and further forestalled their widespread deployment in practical usage. This paper aims to develop explainable DL models for autonomous driving by jointly predicting potential driving actions with corresponding explanations. The explainable DL models can not only boost user trust in autonomy but also serve as a diagnostic approach to identify any model deficiencies or limitations during the system development phase.Design/methodology/approachThis paper proposes an explainable end-to-end autonomous driving system based on “Transformer,” a state-of-the-art self-attention (SA) based model. The model maps visual features from images collected by onboard cameras to guide potential driving actions with corresponding explanations, and aims to achieve soft attention over the image’s global features.FindingsThe results demonstrate the efficacy of the proposed model as it exhibits superior performance (in terms of correct prediction of actions and explanations) compared to the benchmark model by a significant margin with much lower computational cost on a public data set (BDD-OIA). From the ablation studies, the proposed SA module also outperforms other attention mechanisms in feature fusion and can generate meaningful representations for downstream prediction.Originality/valueIn the contexts of situational awareness and driver assistance, the proposed model can perform as a driving alarm system for both human-driven vehicles and autonomous vehicles because it is capable of quickly understanding/characterizing the environment and identifying any infeasible driving actions. In addition, the extra explanation head of the proposed model provides an extra channel for sanity checks to guarantee that the model learns the ideal causal relationships. This provision is critical in the development of autonomous systems. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Intelligent and Connected Vehicles Emerald Publishing

Development and testing of an image transformer for explainable autonomous driving systems

Download PDF

Development and testing of an image transformer for explainable autonomous driving systems

Dong, Jiqian; Chen, Sikai; Miralinaghi, Mohammad; Chen, Tiantian; Labi, Samuel
Journal of Intelligent and Connected Vehicles , Volume 5 (3): 15 – Oct 11, 2022

Abstract

Perception has been identified as the main cause underlying most autonomous vehicle related accidents. As the key technology in perception, deep learning (DL) based computer vision models are generally considered to be black boxes due to poor interpretability. These have exacerbated user distrust and further forestalled their widespread deployment in practical usage. This paper aims to develop explainable DL models for autonomous driving by jointly predicting potential driving actions with corresponding explanations. The explainable DL models can not only boost user trust in autonomy but also serve as a diagnostic approach to identify any model deficiencies or limitations during the system development phase.Design/methodology/approachThis paper proposes an explainable end-to-end autonomous driving system based on “Transformer,” a state-of-the-art self-attention (SA) based model. The model maps visual features from images collected by onboard cameras to guide potential driving actions with corresponding explanations, and aims to achieve soft attention over the image’s global features.FindingsThe results demonstrate the efficacy of the proposed model as it exhibits superior performance (in terms of correct prediction of actions and explanations) compared to the benchmark model by a significant margin with much lower computational cost on a public data set (BDD-OIA). From the ablation studies, the proposed SA module also outperforms other attention mechanisms in feature fusion and can generate meaningful representations for downstream prediction.Originality/valueIn the contexts of situational awareness and driver assistance, the proposed model can perform as a driving alarm system for both human-driven vehicles and autonomous vehicles because it is capable of quickly understanding/characterizing the environment and identifying any infeasible driving actions. In addition, the extra explanation head of the proposed model provides an extra channel for sanity checks to guarantee that the model learns the ideal causal relationships. This provision is critical in the development of autonomous systems.

Loading next page...
 
/lp/emerald-publishing/development-and-testing-of-an-image-transformer-for-explainable-J2RobH8Sr4
Publisher
Emerald Publishing
Copyright
© Jiqian Dong, Sikai Chen, Mohammad Miralinaghi, Tiantian Chen and Samuel Labi. Published in Journal of Intelligent and Connected Vehicles.
ISSN
2399-9802
DOI
10.1108/jicv-06-2022-0021
Publisher site
See Article on Publisher Site

Abstract

Perception has been identified as the main cause underlying most autonomous vehicle related accidents. As the key technology in perception, deep learning (DL) based computer vision models are generally considered to be black boxes due to poor interpretability. These have exacerbated user distrust and further forestalled their widespread deployment in practical usage. This paper aims to develop explainable DL models for autonomous driving by jointly predicting potential driving actions with corresponding explanations. The explainable DL models can not only boost user trust in autonomy but also serve as a diagnostic approach to identify any model deficiencies or limitations during the system development phase.Design/methodology/approachThis paper proposes an explainable end-to-end autonomous driving system based on “Transformer,” a state-of-the-art self-attention (SA) based model. The model maps visual features from images collected by onboard cameras to guide potential driving actions with corresponding explanations, and aims to achieve soft attention over the image’s global features.FindingsThe results demonstrate the efficacy of the proposed model as it exhibits superior performance (in terms of correct prediction of actions and explanations) compared to the benchmark model by a significant margin with much lower computational cost on a public data set (BDD-OIA). From the ablation studies, the proposed SA module also outperforms other attention mechanisms in feature fusion and can generate meaningful representations for downstream prediction.Originality/valueIn the contexts of situational awareness and driver assistance, the proposed model can perform as a driving alarm system for both human-driven vehicles and autonomous vehicles because it is capable of quickly understanding/characterizing the environment and identifying any infeasible driving actions. In addition, the extra explanation head of the proposed model provides an extra channel for sanity checks to guarantee that the model learns the ideal causal relationships. This provision is critical in the development of autonomous systems.

Journal

Journal of Intelligent and Connected Vehicles – Emerald Publishing

Published: Oct 11, 2022

Keywords: Explainable deep learning; Computer vision; Transformer; Autonomous driving

References

  • Why did you predict that? Towards explainable artificial neural networks for travel demand analysis
    Alwosheel, A.; van Cranenburgh, S.; Chorus, C.G.
  • End to end learning for self-driving cars
    Bojarski, M.; Del Testa, D.; Dworakowski, D.; Firner, B.; Flepp, B.; Goyal, P.; Jackel, L.D.
  • Explainable, automated urban interventions to improve pedestrian and vehicle safety
    Bustos, C.; Rhoads, D.; Solé-Ribalta, A.; Masip, D.; Arenas, A.; Lapedriza, A.; Borge-Holthoefer, J.
  • A deep learning algorithm for simulating autonomous driving considering prior knowledge and temporal information
    Chen, S.; Leng, Y.; Labi, S.
  • Graph neural network and reinforcement learning for multi-agent cooperative control of connected autonomous vehicles
    Chen, S.; Dong, J.; Ha, P.; Li, Y.; Labi, S.
  • Traffic graph convolutional recurrent neural network: a deep learning framework for network-scale traffic learning and forecasting
    Cui, Z.; Henrickson, K.; Ke, R.; Wang, Y.
  • Image transformer for explainable autonomous driving system
    Dong, J.; Chen, S.; Zong, S.; Chen, T.; Labi, S.
  • Space-weighted information fusion using deep reinforcement learning: the context of tactical control of lane-changing autonomous vehicles and connectivity range assessment
    Dong, J.; Chen, S.; Li, Y.; Du, R.; Steinfeld, A.; Labi, S.
  • Spatio-weighted information fusion and DRL-based control for connected autonomous vehicles
    Dong, J.; Chen, S.; Li, Y.; Ha, P.Y.J.; Du, R.; Steinfeld, A.; Labi, S.
  • What does explainable AI really mean? A new conceptualization of perspectives
    Doran, D.; Schulz, S.; Besold, T.R.
  • GAQ-EBkSP: a DRL-based urban traffic dynamic rerouting framework using fog-cloud architecture
    Du, R.; Chen, S.; Dong, J.; Ha, P.Y.J.; Labi, S.
  • Evaluation methods and techniques: advanced transportation and congestion management technologies deployment program, tech
  • Vehicle connectivity and automation: a sibling relationship
    Ha, P.; Chen, S.; Du, R.; Dong, J.; Li, Y.; Labi, S.
  • Deep residual learning for image recognition
    He, K.; Zhang, X.; Ren, S.; Sun, J.
  • Vision-based driver assistance systems: survey, taxonomy and advances
    Horgan, J.; Hughes, C.; McDonald, J.; Yogamani, S.
  • Cyber-physical system architecture for automating the mapping of truck loads to bridge behavior using computer vision in connected highway corridors
    Hou, R.; Jeong, S.; Lynch, J.P.; Law, K.H.
  • R-CNN based 3D object detection for autonomous driving
    Hu, H.; Zhao, T.; Wang, Q.; Gao, F.; He, L.
  • Calibrating trust through knowledge: introducing the concept of informed safety for automation in vehicles
    Khastgir, S.; Birrell, S.; Dhadyalla, G.; Jennings, P.
  • Interpretable learning for self-driving cars by visualizing causal attention
    Kim, J.; Canny, J.
  • Advisable learning for self-driving vehicles by internalizing observation-to-action rules
    Kim, J.; Moon, S.; Rohrbach, A.; Darrell, T.; Canny, J.
  • Monocular 3D object detection leveraging accurate proposals and shape reconstruction
    Ku, J.; Pon, A.D.; Waslander, S.L.
  • Platoons of connected vehicles can double throughput in urban roads
    Lioris, J.; Pedarsani, R.; Tascikaraoglu, F.Y.; Varaiya, P.
  • DeepPF: a deep learning based architecture for metro passenger flow prediction
    Liu, Y.; Liu, Z.; Jia, R.
  • Self-driving uber car that hit and killed woman did not recognize that pedestrians jaywalk
    McCausland, P.
  • Collision between vehicle controlled by developmental automated driving system and pedestrian
  • Connected autonomous vehicles for improving mixed traffic efficiency in unsignalized intersections with deep reinforcement learning
    Peng, B.; Keskin, M.F.; Kulcsár, B.; Wymeersch, H.
  • Faster R-CNN: towards real-time object detection with region proposal networks
    Ren, S.; He, K.; Girshick, R.; Sun, J.
  • Capabilities and limitations of peripheral vision
    Rosenholtz, R.
  • MobileNetV2: inverted residuals and linear bottlenecks
    Sandler, M.; Howard, A.; Zhu, M.; Zhmoginov, A.; Chen, L.C.
  • Planning and decision-making for autonomous vehicles
    Schwarting, W.; Alonso-Mora, J.; Rus, D.
  • Transportation decision making: principles of project evaluation and programming, transportation decision making: principles of project evaluation and programming
    Sinha, K.C.; Labi, S.
  • Computer vision based advanced driver assistance system algorithms with optimization techniques-a review
    Sowmya Shree, B.V.; Karthikeyan, A.
  • Exploring applications of deep reinforcement learning for real-world autonomous driving systems
    Talpaert, V.; Sobh, I.; Ravi Kiran, B.; Mannion, P.; Yogamani, S.; El-Sallab, A.; Perez, P.
  • Socioeconomic impacts of automated and connected vehicle: summary of the sixth EU – US
  • TRB forum on preparing for automated vehicles and shared mobility: mini-workshop on the importance and role of connectivity
  • Autonomous vehicle control systems – a review of decision making
    Veres, S.M.; Molnar, L.; Lincoln, N.K.; Morice, C.P.
  • More than the useful field: considering peripheral vision in driving
    Wolfe, B.; Dobres, J.; Rosenholtz, R.; Reimer, B.
  • A framework for the economic evaluation of transport projects, transport notes
  • Periphery-fovea multi-resolution driving model guided by human attention
    Xia, Y.; Kim, J.; Canny, J.; Zipser, K.; Canas-Bajo, T.; Whitney, D.
  • Multi-scale driver behavior modeling based on deep spatial-temporal representation for intelligent vehicles
    Xing, Y.; Lv, C.; Cao, D.; Velenis, E.
  • End-to-end learning of driving models from large-scale video datasets
    Xu, H.; Gao, Y.; Yu, F.; Darrell, T.
  • Explainable object-induced action decision for autonomous vehicles
    Xu, Y.; Yang, X.; Gong, L.; Lin, H.C.; Wu, T.Y.; Li, Y.; Vasconcelos, N.
  • Tesla driver dies in first fatal crash while using autopilot mode
    Yadron, D.; Tynan, D.
  • Forecasting road traffic speeds by considering area-wide spatio-temporal dependencies based on a graph convolutional neural network (GCN)
    Yu, B.; Lee, Y.; Sohn, K.
  • BDD100K: a diverse driving dataset for heterogeneous multitask learning
    Yu, F.; Chen, H.; Wang, X.; Xian, W.; Chen, Y.; Liu, F.; Madhavan, V.; Darrell, T.
  • Exploring self-attention for image recognition
    Zhao, H.; Jia, J.; Koltun, V.
  • Urban flow prediction with spatial–temporal neural ODEs
    Zhou, F.; Li, L.; Zhang, K.; Trajcevski, G.
  • Dynamic prediction of traffic incident duration on urban expressways: a deep learning approach based on LSTM and MLP
    Zhu, W.; Wu, J.; Fu, T.; Wang, J.; Zhang, J.; Shangguan, Q.
  • Automated vision inspection of rail surface cracks: a double-layer data-driven framework
    Zhuang, L.; Wang, L.; Zhang, Z.; Tsui, K.L.