Arguing Machines: Human Supervision of Black Box AI Systems That Make Life-Critical Decisions

Fridman, Lex; Ding, Li; Jenik, Benedikt; Reimer, Bryan

doi:10.1109/cvprw.2019.00173

Cited by 28 publications

(15 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To date, this study has collected more than 15,610 days of data, with over 511,638 miles of driving semi-autonomous vehicles from 122 participants (according to the most recent publication on this work [16]). This study introduces a framework aimed at enhancing semi-autonomous vehicle safety and reliability by building better shared-automated systems [75]. A summary of the previous studies can be viewed in Table 1.…”

Section: Naturalistic Driving Studiesmentioning

confidence: 99%

HARMONY: A Human-Centered Multimodal Driving Study in the Wild

et al. 2021

View full text Add to dashboard Cite

Effective shared autonomy requires a clear understanding of driver's behavior, which is governed by multiple psychophysiological and environmental variables. Disentangling this intricate web of interactions requires understanding the driver's state and behaviors in different real-world scenarios, longitudinally. Naturalistic Driving Studies (NDS) have shown to be an effective approach to understanding the driver's state and behavior in real-world scenarios. However, due to the lack of technological and computing capabilities, former NDS only focused on vision-based approaches, ignoring important psychophysiological factors such as cognition and emotion. The main objective of this paper is to introduce HARMONY, a human-centered multimodal naturalistic driving study, where driver's behaviors and states are monitored through (1) in-cabin and outside video streams (2) physiological signals including driver's heart rate and hand acceleration (IMU data), (3) ambient noise, light, and the vehicle's GPS location, and (4) music logs, including song features such as tempo. HARMONY is the first study that collects long-term naturalistic facial, physiological, and environmental data simultaneously. This paper summarizes HARMONY's goals, framework design, data collection and analysis, and the on-going and future research efforts. Through a presented case study, we first demonstrate the importance of longitudinal driver state sensing through using Kernel Density Estimation Methods. Second, we leverage the application of Bayesian Change Point detection methods to demonstrate how we can identify driver behaviors and responses to the environmental conditions by fusing psychophysiological information with features extracted from video streams.INDEX TERMS Naturalistic driving study, physiological sensing,driver state detection, shared-autonomy, contextual awareness, human-in-the-loop systems

show abstract

Section: Naturalistic Driving Studiesmentioning

confidence: 99%

HARMONY: A Human-Centered Multimodal Driving Study in the Wild

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Tape-based autograd for automatic differentiation. This has facilitated the use of PyTorch deep learning framework for commercial applications by big tech companies such as Tesla for their famous Autopilot system [9] for using in autonomous (self-driving) vehicles and Uber AI Lab for their open-source Pyro [10] which is a probabilistic programming language utilized for deep probabilistic modeling in order to make AI more accessible to the wider community. Our work presented within this paper is based on the latest stable release of the PyTorch framework at the time of writing this paper in order to introduce an improved and an efficient version of CondenseNet [11] CNN for edge devices utilizing self-querying data augmentation and depthwise separable convolutional strategies to improve real-time inference performance as well as reduce the final trained model size, trainable parameters, and FLOPs.…”

Section: Representation Learning In Deep Neural Networkmentioning

confidence: 99%

“…Efficiency and accuracy play a critical role in developing a robust neural network algorithm for computer vision purposes since these algorithms play a crucial role in making life-critical decisions when deployed in real-world use cases such as autonomous (self-driving) cars, for example [9].…”

Section: Recommendationsmentioning

confidence: 99%

EffCNet: An Efficient CondenseNet for Image Classification on NXP BlueBox

Kalgaonkar¹,

El-Sharkawy²

2021

AJECE

View full text Add to dashboard Cite

Intelligent edge devices with built-in processors vary widely in terms of capability and physical form to perform advanced Computer Vision (CV) tasks such as image classification and object detection, for example. With constant advances in the field of autonomous cars and UAVs, embedded systems and mobile devices, there has been an ever-growing demand for extremely efficient Artificial Neural Networks (ANN) for real-time inference on these smart edge devices with constrained computational resources. With unreliable network connections in remote regions and an added complexity of data transmission, it is of an utmost importance to capture and process data locally instead of sending the data to cloud servers for remote processing. Edge devices on the other hand, offer limited processing power due to their inexpensive hardware, and limited cooling and computational resources. In this paper, we propose a novel deep convolutional neural network architecture called EffCNet which is an improved and an efficient version of CondenseNet Convolutional Neural Network (CNN) for edge devices utilizing self-querying data augmentation and depthwise separable convolutional strategies to improve real-time inference performance as well as reduce the final trained model size, trainable parameters, and Floating-Point Operations (FLOPs) of EffCNet CNN. Furthermore, extensive supervised image classification analyses are conducted on two benchmarking datasets: CIFAR-10 and CIFAR-100, to verify real-time inference performance of our proposed CNN. Finally, we deploy these trained weights on NXP BlueBox which is an intelligent edge development platform designed for self-driving vehicles and UAVs, and conclusions will be extrapolated accordingly.

show abstract

“…However, each of the approaches described above iteratively learns to drive a specific circuit -they are not designed to generate an ideal racing line for a new circuit upon which the algorithm has not been previously trained. For real-time application in previously-unseen territory, Fridman et al (2019) trained an end-to-end autonomous driving algorithm using 4.2 million video frames acquired from a Teslahowever, the majority of approaches for roadgoing applications are aimed at safely navigating the road environment, rather than generating a time-optimal path. In any case, end-to-end approaches typically generate the vehicle control inputs from camera footage (Tampuu et al, 2020) rather than a target path from track boundary information.…”

Section: Related Workmentioning

confidence: 99%

Real-Time Optimal Trajectory Planning for Autonomous Vehicles and Lap Time Simulation Using Machine Learning

Garlick¹,

Bradley²

2021

Preprint

View full text Add to dashboard Cite

The widespread development of driverless vehicles has led to the formation of autonomous racing competitions, where the high speeds and fierce rivalry in motorsport provide a testbed to accelerate technology development. A particular challenge for an autonomous vehicle is that of identifying a target trajectory -or in the case of a racing car, the ideal racing line. Many existing approaches to identifying the racing line are either not the time-optimal solutions, or have solution times which are computationally expensive, thus rendering them unsuitable for real-time application using on-board processing hardware. This paper describes a machine learning approach to generating an accurate prediction of the racing line in real-time on desktop processing hardware. The proposed algorithm is a dense feed-forward neural network, trained using a dataset comprising racing lines for a large number of circuits calculated via a traditional optimal control lap time simulation. The network is capable of predicting the racing line with a mean absolute error of ±0.27m, meaning that the accuracy outperforms a human driver, and is comparable to other parts of the autonomous vehicle control system. The system generates predictions within 33ms, making it over 9,000 times faster than traditional methods of finding the optimal racing line. Results suggest that a data-driven approach may therefore be favourable for real-time generation of nearoptimal racing lines than traditional computational methods.

show abstract

Arguing Machines: Human Supervision of Black Box AI Systems That Make Life-Critical Decisions

Cited by 28 publications

References 18 publications

HARMONY: A Human-Centered Multimodal Driving Study in the Wild

HARMONY: A Human-Centered Multimodal Driving Study in the Wild

EffCNet: An Efficient CondenseNet for Image Classification on NXP BlueBox

Real-Time Optimal Trajectory Planning for Autonomous Vehicles and Lap Time Simulation Using Machine Learning

Contact Info

Product

Resources

About