Autonomous Vehicle Motion Models Using Ghost Rules for Habitual Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomously driving vehicles (ADV) do not account for habitual behaviors of pedestrians and cyclists that are not captured by current traffic regulations, leading to potential safety issues when navigating areas with unmarked crossing points or changed traffic rules.
Innovation Solution
A method for an ADV to adapt its driving behavior by using trace data to derive region-specific 'ghost rules' from sensor data, including images and maps, to predict the habitual movements of traffic participants, integrating these into its motion model for trajectory planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If ADV follows current traffic regulations only, then it operates according to official rules, but it cannot anticipate habitual behaviors of traffic participants
Solution Approach 1:
The system performs preliminary data collection and analysis to identify habitual behaviors before the ADV encounters situations where such behaviors may occur. Trace data from sensors is collected and processed to derive ghost rules in advance, allowing the motion model to anticipate habitual behaviors proactively rather than reactively
Solution Approach 2:
The motion model acts as an intermediary layer between the official traffic regulations and the ADV's decision-making process. It integrates both official rules and derived ghost rules describing habitual behaviors, allowing the ADV to consider multiple factors when planning its trajectory and predicting traffic participant actions
2Loss of information
If ADV uses statistical observation of hot spots, then it can identify habitual crossing areas, but the method is time consuming and not applicable to large road networks
Solution Approach 1:
The system performs preliminary data collection and analysis to identify habitual behaviors before the ADV encounters situations where such behaviors may occur. Trace data from sensors is collected and processed to derive ghost rules in advance, allowing the motion model to anticipate habitual behaviors proactively rather than reactively
Solution Approach 2:
Instead of observing actual traffic participants continuously, the system creates simplified representations called ghost rules that capture essential habitual behaviors. These ghost rules are derived from trace data and stored for quick reference, allowing the ADV to access behavioral patterns without performing time-consuming real-time observations
3Loss of information
If ADV collects and processes trace data from sensors, then it can derive ghost rules for habitual behaviors, but the processing complexity increases
Solution Approach 1:
The system extracts only the essential and relevant features from sensor trace data that indicate habitual behaviors, rather than processing all raw sensor information. By focusing on specific patterns in the trace data that signify habitual actions, the system reduces processing complexity while still capturing the necessary behavioral information
Solution Approach 2:
The system transforms raw sensor trace data into a different parameter space representing habitual behaviors through the ghost rules. This parameter transformation simplifies the data structure from complex multi-dimensional sensor readings to standardized behavioral descriptors that are easier to process and integrate into the motion model
Data Source
AI summary
A method for controlling a driving behavior of an autonomously driving vehicle, wherein a processing device performs the following steps for at least one region of a road network: receiving trace data of the region, wherein the trace data describe at least one trace of historic movements and/or behaviors of past traffic participants in the region, wherein the movements and/or behaviors are incompatible with and/or not anticipated by traffic regulations currently valid in the region; deriving rule data describing regular movements and/or behaviors of the past traffic participants described by the trace data; and providing the rule data to a motion model of the vehicle, wherein the vehicle comprises a driving control system (DCS) that plans and/or adapts a driving trajectory by detecting a traffic participant in the region and predicting a future behavior of the participant using the motion model.