以機率模型預測都會區路口的駕駛行為

Probabilistic Modeling of Driver Behaviors at Urban Crossroads

劉員成(Yuan-Cheng Liu)

指導教授 ： 詹魁元

國立臺灣大學/工學院/機械工程學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201903004

摘要

無人車與人類駕駛的互動與溝通，將會在不遠的將來成為主要的議題。在這篇論文中，本論文將研究聚焦於最仰賴互動的無交通號誌十字路口。為了研究駕駛是如何進行決策來安全通過路口，本論文首先定義了路口的重要參數，包括了互動車輛的速度與離路口的距離，並定義了一個駕駛決策行為。藉由研究此決策行為，本論文得到了互動車輛煞車讓道的機率，以及駕駛行為的機率模型。為了驗證此模型，本論文在模擬環境中進行測試，並蒐集真人駕駛的數據進行分析。所得到的驗證結果證明，所提出模型的預測準確率與現有方法接近，且具有更廣泛的應用，同時此預測模型也能夠反映出不同駕駛特徵參數的差異。接著，本論文嘗試使用最佳化方法，藉由所蒐集數據進行駕駛行為的特徵參數回推。儘管此參數辨認的準確率尚有改進空間，目前所得到結果證明了所提出模型在此應用上的可行性。同時，為了驗證此模型在實際路口的可行性，本論文蒐集了真實路口的車輛數據並使用所提出模型進行行為預測，所得結果與模擬環境中的結果一致。最後，本論文將所提出模型應用在無人車的決策行為上並與人類駕駛進行互動，初步結果證明無人車的駕駛行為能夠更順暢的與人類駕駛進行互動。

關鍵字

混流 ； 無人駕駛 ； 機率模型 ； 駕駛行為 ； 十字路口 ； 互動模型 ； 避障

並列摘要

The interactions with human drivers is one of the major challenges for autonomous vehicles in the near future. In this work we consider urban crossroads without signals where driver interactions are indispensable. Crossroad parameters are defined and how drivers passing the crossroad while maintaining a desired speed without collision is studied. A point of action is defined for incoming vehicles from different directions and a probability of yielding for each car is proposed as a function of vehicle speed and the distance-to-intersection for both vehicles. Driver behaviors with these probability models are also proposed. The method is then analyzed and validated by data collected from human drivers in the simulated environments. The result shows comparable prediction accuracy to the state of the art method, where characteristic parameters of drivers are also shown to be critical for the behavior predictions. Afterwards, parameters representing driving styles of drivers are attempted to identify using the optimization approach. In spite of the limited accuracy of parameter identifications, important attributes of the proposed model as well as possible modification are pinpointed. The proposed model is also applied at the urban crossroads to evaluate the applicability in real world. The prediction results are analogous to those acquired in virtual environments. Finally, a procedure is constructed to achieve smoother interactions with human drivers. Preliminary results suggested a human-like computer driver is born while more instances and aspects of evaluations should be accomplished in the future work.

並列關鍵字

mixed-fleet ； autonomous vehicle ； probabilistic model ； driver behavior ； crossroad ； interaction model ； collision avoidance

參考文獻

[1] S. Thrun, M. Montemerlo, H. Dahlkamp, D. Stavens, A. Aron, J. Diebel, P. Fong, J. Gale, M. Halpenny, G. Hoffmann, et al. Stanley: The robot that won the darpa grand challenge. Journal of field Robotics, 23(9):661–692, 2006.

[2] P. Bansal and K. M. Kockelman. Forecasting americans’long-term adoption of connected and autonomous vehicle technologies. Transportation Research Part A: Policy and Practice, 95:49 – 63, 2017.

[3] US Department of Transportation. Preparing for the future of transportation: Automated vehicles 3.0 (av 3.0). Technical report, sep 2018. Accessed: 2019-06-23.

[4] Todd Litman. Autonomous vehicle implementation predictions: Implications for transport planning. Technical report, Victoria Transp. Policy Inst., Victoria, BC, Canada, 2015.

[5] California DMV. Report of traffic collision involving an autonomous vehicle (ol316), 2019.