Search results

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.

This 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.

The 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.

In 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.

This paper aims to propose a new autonomous driving controller to calibrate the absolute heading adaptively. Besides, the second purpose of this paper is to propose a new angle-track loop with a mass regulator to improve the adaptability of the autonomous driving system under different loads and road conditions.

In this paper, the error model of heading is built and a new autonomous driving controller with heading adaptive calibration is designed. The new controller calculates the average lateral error by the self-adjusting interval window and calibrates the absolute heading through the incremental proportional–integral–derivative (PID) controller. A window-size adjustment strategy, based on the current lateral error and the derivative of lateral error, is proposed to improve both the transient and the steady-state responses. An angle-tracking loop with mass regulator is proposed to improve the adaptability of autonomous steering system under different loads and road conditions.

The experiment results demonstrate that this method can compensate the heading installation error and restrain the off-track error from 13.8 to 1.30 cm. The standard error of new controller is smaller than fuzzy-PID calibration controller and the accuracy of autonomous driving system is improved.

The accuracy of heading calibrated by the new controller is not affected by external factors and the efficiency of calibration is improved. As the model parameters of steering system can be obtained manually, the new autonomous steering controller has more simple structure and is easy to implement. Mass regulator is adjusted according to the road conditions and the mass of harvester, which can improve the system adaptability.

Collaborative robotics and autonomous driving are fairly new disciplines, still with a long way to go to achieve goals, set by the research community, manufacturers and users. For technologies like collaborative robotics and autonomous driving, which focus on closing the gap between humans and machines, the physical, psychological and emotional needs of human individuals becoming increasingly important in order to ensure effective and safe human–machine interaction. The authors' goal was to conceptualize ways to combine experience from both fields and transfer artificial intelligence knowledge from one to another. By identifying transferable meta-knowledge, the authors will increase quality of artificial intelligence applications and raise safety and contextual awareness for users and environment in both fields.

First, the authors presented autonomous driving and collaborative robotics and autonomous driving and collaborative robotics' connection to artificial intelligence. The authors continued with advantages and challenges of both fields and identified potential topics for transferrable practices. Topics were divided into three time slots according to expected research timeline.

The identified research opportunities seem manageable in the presented timeline. The authors' expectation was that autonomous driving and collaborative robotics will start moving closer in the following years and even merging in some areas like driverless and humanless transport and logistics.

The authors' findings confirm the latest trends in autonomous driving and collaborative robotics and expand them into new research and collaboration opportunities for the next few years. The authors' research proposal focuses on those that should have the most positive impact to safety, complement, optimize and evolve human capabilities and increase productivity in line with social expectations. Transferring meta-knowledge between fields will increase progress and, in some cases, cut some shortcuts in achieving the aforementioned goals.

The purpose of this research is to achieve multi-task autonomous driving by adjusting the network architecture of the model. Meanwhile, after achieving multi-task autonomous driving, the authors found that the trained neural network model performs poorly in untrained scenarios. Therefore, the authors proposed to improve the transfer efficiency of the model for new scenarios through transfer learning.

First, the authors achieved multi-task autonomous driving by training a model combining convolutional neural network and different structured long short-term memory (LSTM) layers. Second, the authors achieved fast transfer of neural network models in new scenarios by cross-model transfer learning. Finally, the authors combined data collection and data labeling to improve the efficiency of deep learning. Furthermore, the authors verified that the model has good robustness through light and shadow test.

This research achieved road tracking, real-time acceleration–deceleration, obstacle avoidance and left/right sign recognition. The model proposed by the authors (UniBiCLSTM) outperforms the existing models tested with model cars in terms of autonomous driving performance. Furthermore, the CMTL-UniBiCL-RL model trained by the authors through cross-model transfer learning improves the efficiency of model adaptation to new scenarios. Meanwhile, this research proposed an automatic data annotation method, which can save 1/4 of the time for deep learning.

This research provided novel solutions in the achievement of multi-task autonomous driving and neural network model scenario for transfer learning. The experiment was achieved on a single camera with an embedded chip and a scale model car, which is expected to simplify the hardware for autonomous driving.

The anticipated benefits of connected and autonomous vehicles (CAVs) include safety and mobility enhancement. Small headways between successive vehicles, on one hand, can cause increased capacity and throughput and thereby improve overall mobility. On the other hand, small headways can cause vehicle occupant discomfort and unsafety. Therefore, in a CAV environment, it is important to determine appropriate headways that offer a good balance between mobility and user safety/comfort.

In addressing this research question, this study carried out a pilot experiment using a driving simulator equipped with a Level-3 automated driving system, to measure the threshold headways. The Method of Constant Stimuli (MCS) procedure was modified to enable the estimation of two comfort thresholds. The participants (drivers) were placed in three categories (“Cautious,” “Neutral” and “Confident”) and 250 driving tests were carried out for each category. Probit analysis was then used to estimate the threshold headways that differentiate drivers' discomfort and their intention to re-engage the driving tasks.

The results indicate that “Cautious” drivers tend to be more sensitive to the decrease in headways, and therefore exhibit greater propensity to deactivate the automated driving mode under a longer headway relative to other driver groups. Also, there seems to exist no driver discomfort when the CAV maintains headway up to 5%–9% shorter than the headways they typically adopt. Further reduction in headways tends to cause discomfort to drivers and trigger take over control maneuver.

In future studies, the number of observations could be increased further.

The study findings can help guide specification of user-friendly headways specified in the algorithms used for CAV control, by vehicle manufacturers and technology companies. By measuring and learning from a human driver's perception, AV manufacturers can produce personalized AVs to suit the user's preferences regarding headway. Also, the identified headway thresholds could be applied by practitioners and researchers to update highway lane capacities and passenger-car-equivalents in the autonomous mobility era.

The study represents a pioneering effort and preliminary pilot driving simulator experiment to assess the tradeoffs between comfortable headways versus mobility-enhancing headways in an automated driving environment.

Unprecedented endeavors have been made to take autonomous trucks to the open road. This study aims to provide relevant information on autonomous truck technology and to help logistics managers gain insight into assessing optimal shipment sizes for autonomous trucks.

Empirical data of estimated autonomous truck costs are collected to help revise classic, conceptual models of assessing optimal shipment sizes. Numerical experiments are conducted to illustrate the optimal shipment size when varying the autonomous truck technology cost and transportation lead time reduction.

Autonomous truck technology can cost as much as 70% of the price of a truck. Logistics managers using classic models that disregard the additional cost could underestimate the optimal shipment size for autonomous trucks. This study also predicts the possibility of inventory centralization in the supply chain network.

The findings are based on information collected from trade articles and academic journals in the domain of logistics management. Other technical or engineering discussions on autonomous trucks are not included in the literature review.

Logistics managers must consider the latest cost information when deciding on shipment sizes of road freight for autonomous trucks. When the economies of scale in autonomous technology prevail, the classic economic order quantity solution might again suffice as a good approximation for optimal shipment size.

This study shows that some models in the literature might no longer be applicable after the introduction of autonomous trucks. We also develop a new cost expression that is a function of the lead time reduction by adopting autonomous trucks.

The purpose of this paper is to inform researchers who are examining the adoption of autonomous vehicle technology and to provide marketing insights for developers and manufacturers of such vehicles and their ancillary technologies.

This study assesses consumer attitudes and behavioral intentions regarding autonomous vehicles (AV) by applying the consumer version of the unified theory of acceptance and use of technology (UTAUT2). We validate the model through a behavioral research study (n = 1,154).

The findings suggest that attitude toward AV is primarily formed through performance expectancy, effort expectancy, social influence and hedonic motivation. Furthermore, the level of autonomy has limited effects on attitude.

This is the first study to examine attitudes toward AV through the theoretical lens of UTAUT2. Additionally, this study provides insights into consumer perceptions and the corresponding effects on attitude by moderating the level of autonomy.