HADRIAN Research Questions for Collaborating Partners
The HADRIAN project is seeking external collaborations with interested universities on the research questions that are listed below. These questions concern possible improvements of the human driver role in the interaction with automated driving systems. Collaborations will start with these research questions but will be elaborated and refined throughout the collaboration research definition process with a contact within the HADRIAN consortium. Most research questions require a driving simulator though the fidelity and realism of the simulator is open. Collaborations with HADRIAN consortium partners to conduct the studies can be explored as well.
1. To what extent can the quality of take-over maneuvers from Automated Driving Level (ADL) 2 to manual driving be improved by reducing the amount of automation monitoring that the driver needs to engage in while still needing to monitor objects and events in the environment?
As defined in SAE J3016, during ADL 2, the driver has to stay sufficiently engaged in the driving task to take back control from automated driving when needed. For this task, the driver has to monitor the environment to detect and respond to safety critical objects and events (referred to as OEDR for object and event detection and recognition), monitor the vehicle’s performance (e.g. the brake performance) as well as the performance of the driving automation (e.g. that the Lane-Keep-Assistance is engaged and working within expectable limits). However, past research has found that engaging in these monitoring tasks while removing active control tasks can become tedious for drivers and tends to lead to a lack of compliance that has already caused accidents. Therefore, this research topic should investigate to what extent reducing the amount of automation monitoring (while maintaining all other types of monitoring) leads to a decrement or improvement in transition quality from automated to manual driving. Conceptually, here it is assumed that automation monitoring could be automated such that the driver receives a warning with a 5 seconds time prior to the automation disengagement (in contrast to today where this happens often instantaneously). Research to this topic should start with an appropriate theoretical account for the cognitive and perceptive processes that the driver has to engage in and derive an appropriate hypothesis that is then investigated in an experimental study using a driving simulator. It should be considered that the actual implementation of the (prototype) warning system may have an impact on the transition performance and a design that helps the driver in the best way should be considered.
2. To what extent can the quality of take-over maneuvers from Automated Driving Level (ADL) 2 to manual driving be impacted by supporting the driver in adapting the amount of environmental monitoring to the complexity and safety criticality of the environment?
Similar to question 1, this question concerns the automated driving at ADL 2 and only addresses one of the three required monitoring aspects, that is, the monitoring of the environment. The idea is here to help the driver adapt the monitoring to the environment, therefore optimize the monitoring and possible reduce the workload while still maintaining sufficient awareness of the environment. The amount of environmental monitoring that a driver needs to engage in depends on the amount of potential hazards and elements in the environment. Driving on a straight road in the desert likely requires less monitoring than driving in a busy urban street. Therefore, helping the driver adjust his or her monitoring to the environmental complexity, e.g. through an assistant that reminds the driver to monitor more or less, may improve the acceptability of the monitoring, the satisfaction with it, as well as the performance. The question is thereby especially what kind of monitoring interventions could help the driver there (nudging? Warning? Advising? Etc.). Research to this topic should start with an appropriate theoretical account for the cognitive and perceptive processes that the driver has to engage in and derive an appropriate hypothesis that is then investigated in an experimental study using a driving simulator.
3. To what extent can a fixed transition time of 15 sec from automated driving at ADL 3 to manual driving along with “fluid” assistance that adapts to the state and behavior of the user and aids him or her to transition back to manual driving? This fluid assistance should be with a baseline where a minimum of 6 sec transition time at variable rates (e.g. sometimes the transition time may be 10 sec and sometimes only 6 sec minimum)?
In HADRIAN ADL 3 the driver is helped throughout the transition process from automated to manual driving with a scaffolding process, that monitors whether the driver engages in the necessary behavior for a safe transition and intervenes as needed to help the transition. For example, if the driver fails to check the current speed of the vehicle, the software may announce the speed or ask the driver to check the speed. This dynamic aiding process toward a standardized transition sequence is enabled by a guaranteed minimum transition time that again is enabled by appropriate road infrastructure integration (the look-ahead horizon of the vehicle is enlarged through a network of roadside mounted sensors). This standardization of the transition process is not dissimilar to the standard process of transitioning control between pilots and flight automation in aviation. Research to this topic should start with an appropriate theoretical account for the cognitive and perceptive processes of the two different transitioning processes and derive an appropriate hypothesis that is then investigated in an experimental study using a driving simulator.
4. To what extent can a “time-to-transition” that is known to the driver during ADL 3 lead to improved quality of the transition back to manual driving compared to a baseline where this is unknown?
In HADRIAN ADL 3 the user of the automated driving vehicle knows beforehand the duration of non-interrupted periods of automated driving. This will allow the user to better plan his or her non-driving related activities (referred to as NDRA such as watching a movie or working on an Excel file on the tablet). However, this predictability may also allow to improve the quality of the transition back to manual driving because the user can better prepare for ending the NDRA in time and get ready for manual driving. Research to this topic should start with an appropriate theoretical account for the cognitive and perceptive processes of different NDRAs and derive appropriate hypotheses that are then investigated in an experimental study using a driving simulator.
5. To what extent can roadside sensors and C-ITS network (see e.g. https://eur-lex.europa.eu/eli/dec/2017/2380/oj) be used to improve the predictability of automated driving at level 3 for the user in terms of starting the automated driving, non-interrupted execution, and planned transition back to manual driving, as well as extend the automated driving at level 3?
Vehicle mounted sensors have a certain range beyond which the predictability of automated driving operations is limited. An obstacle behind the curve or snow covered lane markings can be a challenge for the automated driving functionality therefore potentially requiring a transition back to manual driving. The numerous occurrence of such transitions back to manual driving can cause acceptance issues because the driver may be unable to sufficiently engage in their non-driving related activities which probably motivated the purchase of often expensive automated driving features. However, even more importantly, the transition back to manual driving represents currently still a safety risk as the driver has to actively understand the situation and set the appropriate actions during the transition. Therefore, it is desirable to reduce the number of necessary transitions or, when necessary, at least make them more predictable. Therefore, the predictability of the transition from automated to manual driving period could be extended through roadside sensors that transmit information to the automated driving vehicle about upcoming unexpected objects, conditions, or events. Furthermore, the same infrastructure could be used to extend the automated driving if critical information is transmitted to the vehicle’s trajectory planner to accommodate the upcoming situation. Therefore, to address this research question an analysis of different applicable scenarios should be conducted and technical solutions within the framework of C-ITS and the HADRIAN consortium identified. Technical solutions should be demonstrated and tested either via simulation or via real-world environment testing.