Solution Concepts in Hierarchical Games Under Bounded Rationality With Applications to Autonomous Driving
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 addresses the challenge of modeling human driving behavior for autonomous vehicle (AV) motion planning by integrating bounded rationality into hierarchical game theory. Traditional game-theoretic approaches assume complete rationality, which often fails to predict human actions accurately. The authors argue that driving is a cognitively demanding, bounded rational activity and propose adapting behavioral game theory models to hierarchical games, a framework that decomposes planning into long-horizon strategic maneuvers and short-horizon tactical trajectories. The primary goal is to identify which solution concepts best fit naturalistic human driving data, thereby improving the predictive capacity of AV planners. The study formalizes a general model of hierarchical games and extends three behavioral models: two based on Quantal Level-k (QLk) and one based on Nash Equilibrium with quantal errors. The authors evaluate these models using a large-scale dataset of human driving at a busy urban intersection, comprising approximately 3,913 agents and 43,765 decision points. The experimental design involves a 2-level hierarchical game where Level 1 consists of discrete maneuvers (e.g., turn or wait) and Level 2 consists of continuous trajectories. To handle the continuous action space, the authors employ three trajectory sampling schemes: single normative sampling (S(1)), bounds sampling (S(1+B)) which includes extreme physical limits, and Gaussian sampling (S(1+G)) which adds stochastic variation around the normative path. Utility functions are constructed based on traffic psychology principles, balancing excitatory motivations (progress toward destination) and inhibitory motivations (safety gaps and pedestrian right-of-way). The results indicate that modeling driving behavior as pure strategy Nash Equilibria with quantal errors at the maneuver level, combined with bounds sampling of actions at the trajectory level, provides the best fit to naturalistic driving data. This specific combination outperformed other configurations in terms of model fit and predictive accuracy. The study also highlights that situational factors significantly impact the performance of behavior models, suggesting that no single model is universally optimal across all traffic contexts. The authors demonstrate that relaxing the assumption of complete rationality allows for more accurate characterization of human error and sub-optimal behavior, which is critical for safe AV interaction. The significance of this work lies in its contribution to the development of safer and more realistic AV motion planners. By providing a formalized framework for hierarchical games under bounded rationality and validating it with a publicly available, large-scale dataset, the paper offers a robust method for estimating behavior model parameters. The findings suggest that AV planners should account for the specific bounded rational strategies humans employ, such as sampling extreme or normative trajectories, rather than assuming perfect rationality. This approach enhances the AV's ability to predict human actions in complex, multi-agent environments, ultimately improving safety and efficiency in mixed traffic scenarios.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-18 |
| archive | success | canonical_url | — | — | 1 | 2026-06-25 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-18 |
| chunk | success | chunk | — | — | 1 | 2026-06-18 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-18 |
| promote | success | — | — | — | 1 | 2026-06-18 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-18 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; 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).
- Theoretical Contribution: computational model, theory or model