Autonomous Driving Ambiguity Resolution via Driver Communication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face limitations in accurately interpreting the intentions of objects in their environment, such as pedestrians, due to the limitations of sensor data and lack of additional information, leading to unreliable decision-making in uncertain traffic situations.
Innovation Solution
A system that includes a communication module to request additional information from the vehicle driver for ambiguous objects, enhancing the environment representation and enabling more accurate control decisions by integrating driver-provided information into the vehicle's control algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the system relies solely on sensor data and environment representation to make driving decisions, then the system maintains high automation level, but the reliability of decision-making deteriorates in uncertain traffic situations
Solution Approach 1:
The communication module serves as an intermediary between the autonomous driving system and the driver. When sensor data is insufficient to reliably determine the intentions of ambiguous objects, the system activates this intermediary to obtain additional information from the driver, thereby maintaining high automation while improving decision-making reliability in uncertain situations.
2Reliability
If the system requests additional information from the driver for every ambiguous object, then the reliability of control decisions improves, but the ease of operation deteriorates due to excessive driver involvement
Solution Approach 1:
The system applies partial action by selectively requesting information from the driver only when sensor data is insufficient and the situation is genuinely ambiguous. The representation generation means classifies objects as ambiguous only when confidence is below a threshold, ensuring that driver involvement occurs only when necessary, thus maintaining reliability without excessive interaction burden.
3Measurement precision
If the system uses sophisticated sensor arrays and environment representation, then the measurement precision of the environment improves, but the loss of information persists due to inherent sensor limitations
Solution Approach 1:
The system merges two information sources: sensor data from sophisticated sensors and human knowledge from the driver. The communication module combines the precise spatial and physical information from sensors with the driver's ability to infer intentions from contextual cues, thereby compensating for the information loss that sensors alone cannot recover.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a system and a corresponding method for autonomously driving a vehicle. Control commands for autonomously or partially autonomously driving a vehicle are generated on the basis of an environment representation which is itself generated from signals of sensing means that observe the environment of the vehicle. In case that ambiguous objects are identified in the environment representation, the system generates an information request for the objects that are identified as being ambiguous and directs it to the driver. Information obtained from the driver is analyzed and additional information on the object is extracted and together with the information derived from the sensing means signal accumulated in the environment representation map. Then on the basis of such enhanced representation map, the traffic situation is determined and suitable control signals are generated for the vehicle actuators. Additionally in case that the system is not capable of deciding on a traffic situation with sufficient confidence, the driver is asked by an information request either to disambiguate the situation or to instruct on how to deal with the ambiguous traffic situation.