Periodic Reporting for period 1 - i4Driving (Integrated 4D driver modelling under uncertainty)
Okres sprawozdawczy: 2022-10-01 do 2024-03-31
The vision of i4Driving is to lay the foundation for a new industry-standard methodology to establish a credible and realistic human road safety baseline for virtual assessment of CCAM systems. The two central ideas we propose are (1) a multi-level, modular and extendable simulation library that combines existing and new models for human driving behavior; in combination with (2) an innovative cross-disciplinary methodology to account for the huge uncertainty in both human behaviors and use case circumstances.
To establish a new safety assessment standard for CCAM relative to human drivers, we argue that we need to address 3 major scientific challenges (CH), which we translate to 9 specific objectives (SO) in this project, all realistically achievable, measurable, and verifiable through defined KPIs. All challenges are highly interdisciplinary, and 9 key innovations (INNO) were envisage to address them, bringing us (far) beyond the state-of-the-art (in research, products and services already available in the market).
1st challenge: A human driver model that captures the relevant behavioural mechanisms for safety assessment. Two specific objectives to address this challenge:
• SO1: Identify causal relationships between external, and human factors, and safety-critical driver behaviors from naturalistic driving studies (NDS) and driving simulation experiments (DSE) data, at the level of specific driving situations (WP1).
• SO2: Make the most of unveiled patterns and existing behavioural and psychological theories to augment existing models with a perception-cognitive layer (WP2).
2nd challenge: Modelling the heterogeneity of human driving behaviours. Three specific objectives to address this challenge:
• SO3: Define a methodology for generating use-cases and simulation scenarios which continuously challenge human drivers in DSE (WP1, WP3).
• SO4: Map heterogeneity of human and external factors into driving performances by means of DSE, in specific driving situations (WP3).
• SO5: Encode driver heterogeneity into probabilistic human behavioural models (WP4).
3rd challenge: Credibility of model-based inferences. Four specific objectives:
• SO6: Set up a “Modelling of the Modelling Process” (WP2, WP4).
• SO7: Define a methodology to validate, or better corroborate human driver models at multiple scales (WP3, WP4, WP5).
• SO8: Evaluate models in target applications (WP6).
• SO9: Ensure transparency, reproducibility and effective communication (WP7 and WP8).
To establish a new safety assessment standard for CCAM relative to human drivers, we argue that we need to address 3 major scientific challenges (CH), which we translate to 9 specific objectives (SO) in this project, all realistically achievable, measurable, and verifiable through defined KPIs. All challenges are highly interdisciplinary, and 9 key innovations (INNO) were envisage to address them, bringing us (far) beyond the state-of-the-art (in research, products and services already available in the market).
1st challenge: A human driver model that captures the relevant behavioural mechanisms for safety assessment. Two specific objectives to address this challenge:
• SO1: Identify causal relationships between external, and human factors, and safety-critical driver behaviors from naturalistic driving studies (NDS) and driving simulation experiments (DSE) data, at the level of specific driving situations (WP1).
• SO2: Make the most of unveiled patterns and existing behavioural and psychological theories to augment existing models with a perception-cognitive layer (WP2).
2nd challenge: Modelling the heterogeneity of human driving behaviours. Three specific objectives to address this challenge:
• SO3: Define a methodology for generating use-cases and simulation scenarios which continuously challenge human drivers in DSE (WP1, WP3).
• SO4: Map heterogeneity of human and external factors into driving performances by means of DSE, in specific driving situations (WP3).
• SO5: Encode driver heterogeneity into probabilistic human behavioural models (WP4).
3rd challenge: Credibility of model-based inferences. Four specific objectives:
• SO6: Set up a “Modelling of the Modelling Process” (WP2, WP4).
• SO7: Define a methodology to validate, or better corroborate human driver models at multiple scales (WP3, WP4, WP5).
• SO8: Evaluate models in target applications (WP6).
• SO9: Ensure transparency, reproducibility and effective communication (WP7 and WP8).
Summary of progress towards the achievement of each of the project objectives
From month 1 – 18, most of the work on the tasks within the various work packages has started. The i4Driving team has started with the first, ‘inner’ model development cycle. In the inner model development process existing datasets from different sources have been harmonised (in terms of semantics and format) and used for model development. State-of-the-art data mining techniques were developed (INNO1), to unveil patterns and formulate plausible hypotheses in these data related to human (and external) factors and driving behaviours, to identify model requirements (WP1). This inner model development process moves from conceptualisation to calibration and validation, through computer code implementation (WP2).
The hypotheses resulting from INNO1 and the new methodologies to identify relevant use-cases and safety critical scenarios that have been developed and automatically generating the critical driving situations in driving simulators (INNO3 - 4).
The design of experiments on driving simulators and test tracks has started and it is now finished for the simulation experiments (WP3), and concluded for one simulator, ongoing on another one and will start soon in the other 2 simulators.
The design of applications for i4Drivng model (WP6) has also started and the alignment between WP2, WP3, WP4, WP5 and WP6 has been discussed and aligned.
As mentioned above, this first model cycle uses the existing datasets and research developed so far, addressing mainly the specific objectives 1-5. In the second model cycle, the data from the simulation experiments and test tracks experiments will be use to better calibrate i4Driving model, reaching the conclusion of the specific objectives.
Significant activities in support of these achievements
In the first 18 months of the project, the main activities focused on the initial model development cycle, leveraging existing evidence and research. Here's a detailed breakdown of the activities:
• Collection and Processing of Naturalistic Driving and Traffic Datasets (WP1 & WP4)
• Identification and Coding of Human Perception and Cognitive Processes (WP1 & WP2)
• Definition and Coding of Methods to Manage Modelling Uncertainty (WP2 & WP4)
• Design of Experiments on Driving Simulators and Test Tracks (WP3 & WP5)
• Design of Applications for i4Driving Models (WP6)
These activities are critical for laying the groundwork for advanced model development, ensuring that the project is built on a solid foundation of empirical data and well-defined cognitive and perceptual processes.
From month 1 – 18, most of the work on the tasks within the various work packages has started. The i4Driving team has started with the first, ‘inner’ model development cycle. In the inner model development process existing datasets from different sources have been harmonised (in terms of semantics and format) and used for model development. State-of-the-art data mining techniques were developed (INNO1), to unveil patterns and formulate plausible hypotheses in these data related to human (and external) factors and driving behaviours, to identify model requirements (WP1). This inner model development process moves from conceptualisation to calibration and validation, through computer code implementation (WP2).
The hypotheses resulting from INNO1 and the new methodologies to identify relevant use-cases and safety critical scenarios that have been developed and automatically generating the critical driving situations in driving simulators (INNO3 - 4).
The design of experiments on driving simulators and test tracks has started and it is now finished for the simulation experiments (WP3), and concluded for one simulator, ongoing on another one and will start soon in the other 2 simulators.
The design of applications for i4Drivng model (WP6) has also started and the alignment between WP2, WP3, WP4, WP5 and WP6 has been discussed and aligned.
As mentioned above, this first model cycle uses the existing datasets and research developed so far, addressing mainly the specific objectives 1-5. In the second model cycle, the data from the simulation experiments and test tracks experiments will be use to better calibrate i4Driving model, reaching the conclusion of the specific objectives.
Significant activities in support of these achievements
In the first 18 months of the project, the main activities focused on the initial model development cycle, leveraging existing evidence and research. Here's a detailed breakdown of the activities:
• Collection and Processing of Naturalistic Driving and Traffic Datasets (WP1 & WP4)
• Identification and Coding of Human Perception and Cognitive Processes (WP1 & WP2)
• Definition and Coding of Methods to Manage Modelling Uncertainty (WP2 & WP4)
• Design of Experiments on Driving Simulators and Test Tracks (WP3 & WP5)
• Design of Applications for i4Driving Models (WP6)
These activities are critical for laying the groundwork for advanced model development, ensuring that the project is built on a solid foundation of empirical data and well-defined cognitive and perceptual processes.