| 研究課題/領域番号 |
23KJ0391
|
|---|
| 研究機関 | 東京大学 |
|---|
研究代表者 |
林 鵬飛 東京大学, 情報理工学系研究科, 特別研究員(DC2)
|
|---|
| 研究期間 (年度) |
2023-04-25 – 2025-03-31
|
| キーワード | Autonomous Driving / Collision Avoidance / Path Planning / Model Predictive Control / Optimization |
|---|
| 研究実績の概要 |
Since 2023, I have four first-authored papers that are accepted in prestigious conferences and journals, showcasing my contributions to autonomous driving technologies. These publications include innovative strategies for occlusion-aware path planning, interactive speed optimization in potential field-based path planning, and advanced lane-changing tactics considering time-to-collision. Notably, I've contributed to developing strategies for lane-changing, which reduced both maneuver length and path curvature by 27.1% and 56.1%, respectively, thus improving driving efficiency and passenger comfort. My work in occlusion-aware path planning has provided more effective solutions for unexpected vehicle intrusions, enhancing overall road safety.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
I have currently completed about half of the intended progress. However, I have encountered computational power limitations in the aspect of Virtual Scene Construction. This challenge arises mainly due to the high complexity and resource-intensive nature of accurately simulating and rendering real-world traffic scenarios in virtual environments. Tools like SUMO and CARLA require significant processing capabilities to simulate intricate vehicle dynamics and environmental conditions effectively. Despite optimization efforts, the current computational resources are proving to be a bottleneck, hindering the seamless integration of dynamic potential fields with real-time virtual scenario construction and thereby impacting the overall progression of the research plan.
|
| 今後の研究の推進方策 |
My future research direction will focus on enhancing autonomous vehicle (AV) systems by integrating various components for comprehensive decision-making and action.I am going to build a rule-adherence decision making that uses responsibility-sensitive safety (RSS) to ensure that decisions are made following traffic rules and ethical standards. It involves using a Reinforcement Learning (RL) agent and input from a human driving expert to refine the policy network for smarter and safer decision-making.
|
| 次年度使用額が生じた理由 |
The consolidated funds will be strategically utilized for a month-long research exchange visit to foster collaboration and gain new insights, which is essential for the advancement of my research. Additionally, a portion of the budget will be allocated for professional English language editing and publication fees for my journal papers, ensuring that the results of my work are communicated effectively and meet the high standards required for international dissemination. This plan aligns with my commitment to enhancing academic excellence and contributing to the global scientific community.
|
