Driving Route Generation for Real-World Emissions Compliance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods struggle to predict and optimize emissions of internal combustion engine vehicles across various driving cycles, leading to potential non-compliance with emission standards and inefficient emissions management, as they often focus on critical cycles at the expense of frequent, less critical ones.
Innovation Solution
A computer-implemented method using a machine learning system to generate representative driving histories and speed profiles that mimic real-world driving conditions, employing a generator and discriminator trained with random vectors and additional vehicle and driver characteristics, optimizing emissions by minimizing differences in distribution between generated and actual driving data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If emission optimization is focused on critical driving cycles, then compliance with emission standards in challenging cycles is improved, but overall emissions in real-world operation deteriorate due to neglect of frequent less critical cycles
Solution Approach 1:
The patent transforms the evaluation criterion from binary compliance (pass/fail) to a continuous distribution-based assessment. By comparing the distribution characteristics of generated driving cycles against reference cycles using statistical parameters (mean, standard deviation, skewness, kurtosis), the system identifies optimization opportunities across the entire driving cycle spectrum rather than focusing solely on critical cases.
Solution Approach 2:
The patent generates synthetic driving cycles that replicate the statistical distribution characteristics of real-world driving patterns. By creating virtual driving cycles with matched distribution parameters, the system enables comprehensive emission assessment without requiring exhaustive testing of all possible driving scenarios, thus addressing both compliance and overall emission reduction.
2Reliability
If comprehensive testing across all conceivable driving cycles is performed, then emission compliance reliability is improved, but assessment complexity and resource requirements increase significantly
Solution Approach 1:
The patent creates simplified virtual representations (copies) of complex real-world driving cycles by matching key distribution parameters. These synthetic driving cycles capture the essential statistical characteristics of actual driving patterns while being computationally tractable, enabling comprehensive emission assessment without the complexity of simulating every possible real-world scenario.
Solution Approach 2:
The patent reduces the complexity of comprehensive driving cycle assessment by transforming the problem from evaluating infinite continuous driving patterns to analyzing a finite set of distribution parameters (mean, standard deviation, skewness, kurtosis). This parameter transformation enables systematic evaluation while maintaining assessment reliability.
3Adaptability or versatility
If driving cycle selection is left to inspector discretion, then flexibility in identifying critical cases is improved, but consistency and objectivity of emission assessment deteriorate
Solution Approach 1:
The patent creates a universal assessment framework that simultaneously provides inspector flexibility in generating custom driving cycles and ensures objective consistency through automated distribution-based evaluation. The system can accommodate various inspector preferences while maintaining standardized assessment criteria through statistical parameter comparison.
Solution Approach 2:
The patent implements an automated feedback mechanism that evaluates driving cycles based on their distribution characteristics rather than inspector subjectivity. The system provides quantitative feedback on how generated driving cycles compare to reference distributions, ensuring consistent and objective emission assessments across different inspectors and applications.
Data Source
Figure 1
AI summary
A computer-implemented method for training a machine learning system to generate driving profiles and/or routes of a vehicle is presented, comprising the following steps: - a generator (41) of the machine learning system (4) receives initial random vectors as input and generates initial driving routes and associated initial driving profiles for each of the initial random vectors, - driving routes and their corresponding driving profiles recorded during operation are stored in a database (2), - second driving routes and their corresponding second driving profiles recorded during operation are selected from the database (2), - a discriminator (42) of the machine learning system (4) receives as inputs initial pairs of initial generated driving routes and their corresponding first generated driving profiles, and second pairs of second driving routes and their corresponding second driving profiles recorded during operation.- The discriminator (42) calculates outputs based on the input variables, which characterize for each pair received as input variables whether it is a first pair consisting of a first generated route and its corresponding first generated journey profile, or a second pair consisting of a second route and its corresponding second journey profile recorded during operation. - Depending on the outputs of the discriminator (42), an objective function (5) is optimized, which represents a distance between the distribution of the first pairs and the distribution of the second pairs.