Adaptive Game-Theoretic Decision Making for Autonomous Vehicle Control at Roundabouts
DOI: 10.48550/arxiv.1810.00829
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 autonomous vehicle (AV) control at unsignalized roundabout intersections, where vehicles must negotiate right-of-way without centralized traffic signals. The authors propose an adaptive game-theoretic decision-making algorithm that models interactions between an ego vehicle and an opponent vehicle. The approach is motivated by the need to account for strategic interactions in multi-vehicle scenarios and to adapt to varying human driving behaviors, specifically distinguishing between conservative and aggressive driver types. The methodology employs a hierarchical control architecture focusing on high-level decision-making using discrete-time vehicle kinematics. The core of the algorithm is a receding-horizon optimal control problem where the ego vehicle maximizes a cumulative reward function defined by six features: collision avoidance, on-road status, distance to objective, safe separation, lane marking compliance, and speed. To predict the opponent’s actions, the authors utilize a level-k game theory framework, simplifying it into two driver types: Type-1 (conservative), who assumes the opponent treats them as a stationary obstacle, and Type-2 (aggressive), who assumes the opponent is a Type-1 driver. The AV controller adaptively estimates the opponent’s type online by comparing predicted actions against actual observed actions, updating a probability belief using a Bayesian-like update rule. To ensure real-time feasibility, the computationally intensive optimization problems are solved offline, and the resulting policies are approximated using three neural networks: one to identify critical states where driver types yield different actions, one to infer the driver type from state and action data, and one to output the optimal ego vehicle action. Simulation results demonstrate the controller’s performance against both modeled Type-1 and Type-2 opponents, as well as human operators controlling the opponent vehicle via keyboard. The neural networks achieved validation accuracies between 96.7% and 98.2%. When interacting with a Type-1 (conservative) opponent, the AV correctly predicted the yield behavior and proceeded through the roundabout first. Conversely, when facing a Type-2 (aggressive) opponent, the AV adapted by decelerating, yielding the right-of-way, and accelerating only after the opponent passed. The simulations confirmed that the adaptive strategy successfully identified the opponent’s driver type over time and adjusted the ego vehicle’s behavior accordingly to maintain safety and efficiency. The significance of this work lies in providing a robust, real-time capable framework for AV decision-making in complex, unsignalized intersections. By integrating game-theoretic modeling with adaptive learning and function approximation, the approach allows autonomous systems to handle diverse human driving styles without requiring explicit prior knowledge of the opponent’s behavior. This contributes to the development of safer and more socially compliant autonomous vehicles capable of navigating interactive traffic scenarios effectively.
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 | OpenAlex-citations | — | — | 1 | 2026-06-18 |
| archive | success | openalex | — | — | 5 | 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