Publication

Road users may need additional information to a vehicle’s speed and position in lane to understand the behavior and intentions of self-driving vehicles, such as external Human Machine Interface (eHMI), i.e. visual signals (lights) indicating the vehicle's status, behavior and intentions, especially in environments where self-driving vehicles are expected to drive at low speeds, for example, in urban environments. The study in this report developed and implemented self-driving functions and control room functions in an automated delivery vehicle (ADV), as well as developed an eHMI concept to communicate the vehicle’s states, intentions, and behaviour to the surroundings. The software stack and the development of the main features of the self-driving driving capabilities including the lateral controller are described in this report. Further, modules such as the Representational State Transfer (REST) for communication, the remote control of the ADV and the eHMI communication interface with the vehicle signals are presented. A lesson learned from the study is that further refinement in repeatability of the initial conditions of the system is essential. Most of the individual parts of the chain from a command to the ADV were created via the user interface in the Autonomous Transport Management System (ATMS). However, the whole chain was hard to achieve, and the need of frequent testing and integration was evident and faults in the chain could result in cumbersome and long procedures to restart the integration test. The study also revealed issues with the stability of the entire system. Several eHMI concepts have been developed in industry and in research. However, up to this date there are no standards or established frameworks for the design of eHMI. Nevertheless, guidelines and recommendations have been proposed in different studies, for example that eHMI should be consistent with existing eHMI, address road users in general and not tell or instruct other road users what to do. The eHMI-concept developed in this study conveyed the following messages: Automation mode, Acceleration, Deceleration and Delivery mode. A model for an eHMI-strategy is also proposed. The eHMI prototype on the ADV in this study was composed of LED lights with multiple color options, an ECU with CAN hardware and software that controlled the eHMI. The initial idea was to use vehicle data from the CAN bus, such as speed and steering angle and the control algorithm worked technically, but the eHMI for acceleration and deceleration were activated/inactivated too fast (within 1-2 seconds) for an observer to perceive and grasp the meaning of the eHMI. A lesson learned was that the activation/inactivation of the eHMI should, therefore, be executed by the computer that manages the autonomous driving functions in the ADV.