Towards Autonomous Cruising on Highways

A Systematic Literature Review About the Impact of Artificial Intelligence on Autonomous Vehicle Safety

“…Considering some specificities of autonomous truck and its risks, at least a few more studies about the topic could be expected. [64], [67], [70], [61], [65], [69], [62], [58], [63], [71], [72], [74], [68], [ [39], [18], [32], [31], [33], [26], [28], [30], [55], [52], [29], [ [75], [41], [29], [20], [44], [35], Prediction of adc vanced driver assistance systems (ADAS) remaining useful life (RUL) for the prognosis of ADAS safety critical components Pedestrian Detection; How to "automate" manual annotation for images to train visual perception for AVs Road junction detection; [52], [27], [37], [30] Bayesian Artificial Intelligence…”

“…The studies on human factor cover important aspects to be considered in the autonomous cars engineering due to the human-in-the-loop factor, such as: safety, comfort, and stability based on the human driver perception behavior [56]; design of real time transition from assisted driving to automated driving under conditions of high probability of a collision [57]; diagnosing and predicting stress and fatigue of driver in semi-automated vehicles [58]; and advances in driver-vehicle interface [59]. Considering the studies (4) proposing conceptual models and frameworks, they have a considerable diversity of focus, such as: ML and cloud-based framework proposed to address safety and reliability-related issues [60]; AV conceptual model [61]; an interdisciplinary framework to extract knowledge from the large amount of available data during driving to reduce driver's behavioral uncertainties [62]; and a proposition of an AV highway concept to improve highway driving safety [63].…”

“…The papers proposing conceptual models and frameworks did not present findings related to experimental results. Most of them relied on general AI/ML instead of a specific technique [61][62] [63]. Also, besides the proposed approaches themselves, they focused the discussions around the issues they aimed to address, the theoretical background and future potential problems to be addressed in the field.…”

“…The same grouping strategy applied to Table V (system-oriented and componentoriented) can be applied here to evaluate the problems (CT.4). System-level problems include 16 papers: AV Validation [64], Machine-learning-based systems validation to the ultradependable levels required for AV [67], Human and Machine Driver Co-existence [60], Coexistence Human Machine Controller [70], Driving Car Tasks Classification [61], Lack of efficient Safety Performance Verification technique when AI/ML is used [65], Crash assignment, especially between automated vehicles and non-automated vehicles [69], Reduce the uncertainty of a driver behavior prediction model [62], Investigate three underexplored themes for AV research: safety, interpretability, and compliance [73], How vehicle autonomy technology can be used to benefit car drivers and also to propose a concept of an autonomous highway vehicle which improves highway driving safety [63], AV decisions in complex dilemmas as a social agent [71], Hybrid (humans and machines) collective decision-making systems [72], Autonomy assurance and trust in Automated Transportation Systems [74], AV Test [66] and, Evaluate the impacts of the number of highly automated vehicles on future traffic safety and traffic flow [68]. Considering the component-level problems, 21 papers (36%) are related to dealing with algorithms and techniques to deal with Machine Perception issues, such as: Vehicle Cyber Attack [39], Turn Signal Recognition [18], Securing connected vehicles against Denial of Service (DoS) attack [51], Road Detection [75], Traffic Light Detection [35], Prediction of advanced driver assistance systems (ADAS) remaining useful life (RUL) for the prognosis of ADAS safety critical components [52], Vehicle Detection and Counting [36], predicts the CPU usage patterns of software tasks running on a self-driving car [53], a safety warning and driver-assistance system and an automatic pilot for rural and urban traffic environments [21], reliable and robust obstacles detection continues to be largely investigated and still remains an open challenge, especially for difficult scenarios and, in general cases, with loosened constraints and multiple simultaneous usecases…”

“…Considering some specificities of autonomous truck and its risks, at least a few more studies about the topic could be expected. [64], [67], [70], [61], [65], [69], [62], [58], [63], [71], [72], [74], [68], [19] Artificial Neural Networks (ANN) [39], [18], [32], [31], [33], [26], [28], [30], [55], [52], [29], [60] Support Vector Machine (SVM) 10 17%…”