ISSN 1671-3710
CN 11-4766/R

Advances in Psychological Science ›› 2021, Vol. 29 ›› Issue (11): 1979-1992.doi: 10.3724/SP.J.1042.2021.01979

• Regular Articles • Previous Articles     Next Articles

Evaluation of external HMI in autonomous vehicles based on pedestrian road crossing decision-making model

JIANG Qianni, ZHUANG Xiangling(), MA Guojie   

  1. Shaanxi Key Laboratory of Behavior and Cognitive Neuroscience, School of Psychology, Shaanxi Normal University, Xi’an 710062, China
  • Received:2020-12-04 Online:2021-11-15 Published:2021-09-23
  • Contact: ZHUANG Xiangling


For autonomous vehicles driven in road context with pedestrians, it is essential to ensure safe and efficient interaction with pedestrians. Compared with the interaction between traditional vehicles and pedestrians, the interaction between autonomous vehicles and pedestrians brings new risks. On the one hand, the driver’s attention may be distracted from the driving task, which resulted in lower reliability of their non-verbal cues. On the other hand, pedestrians are not familiar with autonomous vehicles, which may lead to false expectations of vehicle behavior that led to a higher risk of conflicts. To solve the problem, autonomous vehicles of high level (e.g. above L3) are usually equipped with an external human-machine interface (eHMIs) to communicate with pedestrians.
An overview of current studies shows that the current external eHMIs mainly conveyed vehicle status (whether it is in auto mode), intentions, and road-crossing advice to pedestrians in visual modalities such as text, icon, projection, etc. These eHMIs have been evaluated to determine their effect on pedestrian crossing intention, speed, and accuracy in real as well as simulated contexts.
However, a user-centered design of eHMIs should systematically support pedestrian information processing needs during road crossing decision making. To fill the gap, a conceptual model was proposed to capture pedestrian’s dynamic road crossing decision-making when interacting with autonomous vehicles. The model integrated pedestrian’s road crossing decision-making process and the situation awareness theory.
Based on the model, eHMIs should promote pedestrians’ perception of the traffic elements related with the vehicle, comprehension of their meaning, and the projection of the vehicle’s future behaviors. Design of eHMIs should support pedestrian information processing needs for each of the three phases.
The first phase is to perceive the status and dynamics of vehicles in the traffic environment. To support the perception of the displayed information, the recognizability of information is the key to improve the effectiveness of interfaces. Researchers should apply multiple modalities’ interfaces to convey the vehicle’s information, for example, conveying information by the combination of projection, dynamic light, and icon interface can improve the recognizability. And they should reside interfaces in conjunction with the vehicle, street infrastructure, and the pedestrian to cope with the more complex traffic situation.
The second phase is to comprehend the situation based on information collected in the perception stage. To improve comprehensibility, text interfaces should present concise phrases, and non-text interfaces should be standardized and explained with texts, otherwise, they should be trained to pedestrians to improve comprehensibility. Besides, the perspective of the message also affects the clarity of the displayed information. For red light, it can be interpreted from the perspective of the pedestrian as “Please stop” or from the perspective of the vehicle “I will stop”. An appropriate message perspective can improve pedestrians’ understanding and acceptance of the information so that they can make safe crossing decisions.
In the projection phase, eHMIs need to help pedestrians predict crossing risks and assist them in making decisions quickly and accurately. Researchers should combine eHMIs with vehicle motion information to convey the vehicle’s future action intentions more intuitively. For example, pedestrians can predict vehicle intention more quickly and accurately by presenting the real-time vehicle speed on eHMIs.
For future research, we suggest an extension of current findings to more complex contexts beyond the one-vehicle-one-pedestrian scenario and focus on how the design affects pedestrians in multi-pedestrian and multi-vehicle mixed traffic conditions. Efforts are also needed to understand how the communication interface affected the formation and update of pedestrian situation awareness, as well as the role of mental models in human-vehicle interactions. These mechanisms can facilitate the model-based evaluation of future interfaces and inform new theory-based designs in complex scenarios.

Key words: autonomous vehicles, external human machine interfaces, eHMIs, pedestrian road crossing decision-making model, pedestrian safety

CLC Number: