HMIway-env: A Framework for Simulating Behaviors and Preferences to Support Human-AI Teaming in Driving
DOI: 10.1109/cvprw56347.2022.00480
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper introduces HMIway-env, a lightweight simulation and modeling framework designed to study human-AI teaming in driving contexts. The primary motivation is the need for adaptive AI systems that can respond to individual driver states, traits, and preferences, rather than employing generic assistance. Existing simulators often lack the ability to model internal human states or preferences for receiving help. HMIway-env addresses this by extending the OpenAI Gym-based `highway-env` simulator to incorporate personalized driver behavior models, mechanisms for AI intervention, and models for intervention efficacy. The framework models the driving scenario as a reinforcement learning problem with a joint policy for both the human driver and the AI system. The authors introduce a `PilotedMDPVehicle` class that encapsulates a distraction model and an intervention acceptance model. The distraction state is modeled as a Markov chain, where transition probabilities are modulated by the driver’s willingness to accept AI alerts. Key parameters include an "obstacle inflation factor" to represent driver cautiousness and an "intervention effectiveness factor" to represent receptiveness to alerts. The AI agent issues binary alerts, which influence the driver’s distraction state over a fixed time window. The reward structure incentivizes safe driving, high speed, and sparse, effective alerts. The authors validate the framework through computational experiments and a human subjects study. In computational roll-outs, they compared two driver types: one less cautious and unreceptive to alerts, and another more cautious and receptive. Results showed that the AI could effectively assist the receptive driver, maintaining higher performance metrics and allowing for closer following distances (lower Time-To-Collision) due to increased confidence in the joint team. The AI interventions were more active when driving rewards decreased, particularly for the unreceptive driver. To validate the realism of the simulated behaviors, the authors conducted a crowd-sourced study with 500 participants. Subjects viewed videos of simulated driving with varying levels of cautiousness and distraction and rated the perceived risk and distraction. The results indicated that participants’ intuitions about riskiness aligned with the model parameters; specifically, perceived risk increased as the distraction parameter increased. This validation supports the framework’s utility as a data-generation engine for training personalized human-AI teaming policies, demonstrating that it can faithfully simulate complex human behaviors like distraction and varying levels of caution.
Key finding
Crowd-sourced validation confirmed that the simulated driving behaviors generated by the HMIway-env framework accurately matched human intuitions regarding risk and distraction levels.
Methodology
mixed_methods
Sample size: 500
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via author_sweep_intake on 2026-05-28.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 2 | 2026-05-28 |
| archive | success | canonical_url | — | — | 1 | 2026-06-06 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-04 |
| chunk | success | chunk | — | — | 1 | 2026-06-04 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-04 |
| enrich | success | — | — | — | 1 | 2026-05-28 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
Information type
What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).
- Methodological Resource: tool software
- Theoretical Contribution: computational model, conceptual framework