Automated Vehicle eHMI for Pedestrian Interaction Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in controlling external human-machine interfaces (eHMI) of automated vehicles to enhance interaction with pedestrians in dynamic traffic environments is often overlooked, leading to uncertainty and inefficiency in decision-making processes.
Innovation Solution
A method and system that utilize sensor information and behavior planning algorithms to predict pedestrian actions, plan vehicle behaviors, and communicate intentions through an external human-machine interface (eHMI) to ensure clear and cooperative interactions, reducing uncertainty and improving decision-making.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If external human machine interface (eHMI) is used to communicate vehicle intentions to pedestrians, then interaction clarity and cooperative behavior are improved, but system complexity and decision-making computational load increase
Solution Approach 1:
The system performs preliminary prediction of pedestrian behaviors and future trajectories before making final interaction decisions. By pre-computing multiple potential pedestrian actions and their corresponding optimal vehicle responses, the system prepares communication signals in advance, reducing real-time computational complexity while maintaining interaction clarity.
Solution Approach 2:
The eHMI system acts as an intermediary communication channel between the automated vehicle and pedestrians. It translates complex internal decision-making processes into simple, understandable visual signals that convey vehicle intentions to pedestrians, improving interaction clarity without requiring pedestrians to understand the complex system behind the decisions.
2Reliability
If multiple predicted pedestrian behaviors are considered in behavior planning, then interaction safety and cooperative resolution are improved, but computational time and processing complexity increase
Solution Approach 1:
The system computes a limited set of the most probable pedestrian behaviors rather than all possible behaviors. By focusing computational resources on predicting only the most likely interaction scenarios based on pedestrian state and context, the system achieves sufficient interaction safety without the excessive computational time required to model every possible pedestrian action.
Solution Approach 2:
The system pre-identifies and prioritizes the most likely pedestrian behaviors based on initial sensor data and pedestrian state. This preliminary filtering of behavior options before detailed trajectory prediction reduces the computational burden while maintaining safety by ensuring the most critical scenarios are thoroughly analyzed.
3Loss of information
If eHMI communicates detailed behavior information to pedestrians, then transparency and trust are improved, but information overload and pedestrian confusion may occur
Solution Approach 1:
The system extracts and communicates only the most relevant aspects of vehicle behavior intentions to pedestrians through eHMI, rather than transmitting all available information. By selectively presenting key information such as primary intended action and confidence level, the system maintains transparency while avoiding information overload that would confuse pedestrians.
Solution Approach 2:
The eHMI communication adapts its information content and detail level based on the specific interaction context and pedestrian state. Different situations receive different levels of information detail - for example, high-confidence predictable scenarios receive simpler signals while uncertain or complex scenarios receive more detailed information, optimizing both transparency and ease of understanding for each local context.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and system for assisting an ego-agent in operating in a dynamic environment in which at least one other agent is. By applying a behavior-planning algorithm based on acquired sensor information and agent information, ego-behaviors of the ego-agent are planned. Furthermore, behaviors of the at least one other agent are predicted. The method determines an interaction zone in the environment of the ego-agent and at least one other agent based on the planned ego-behaviors and the predicted behaviors of the at least one other agent. A behavior planner selects a combination of one planned ego-behavior and one corresponding predicted behavior based on a predicted interaction of the ego-agent with the at least one other agent in the interaction zone, and the ego-agent is operated based on the selected ego-behavior. The method further comprises deciding whether to output a signal to the at least one other agent based on fulfilling at least one set of criteria and based on the selected combination of one planned ego-behavior and one corresponding predicted behavior, generating the signal including information on the selected combination behavior of the ego-agent and the predicted behavior of the at least one other agent, and, in case of deciding to output the generated signal, outputting the signal via an external human machine interface to the at least one other agent.