基于改进型Stackelberg博弈的自动驾驶测试数据定价模型

Pricing Model of Autonomous Vehicle Testing Data Based on Evolved Stackelberg Game

摘要:

针对自动驾驶测试数据兼具连续与离散变量，且包含时间戳和经纬度等间接信息特征的特点，利用特征挖掘过滤、连续变量离散化、驾驶模式加权等方法对传统信息熵方法进行适应性调整，提出基于特征工程的驾驶模式加权信息熵方法，确定自动驾驶测试数据信息量；引入信息量构建数据消费者效用方程，提出考虑信息量和平台利润率约束的改进型Stackelberg博弈数据定价模型。以上海市自动驾驶实际测试数据开展典型案例分析，结果表明，基于改进型Stackelberg博弈的数据定价模型可有效评估数据信息量，合理分配数据生产者、数据平台和数据消费者交易三方的利润率，并显著提升数据交易量和数据交易三方总效用，从而增强自动驾驶测试数据交易市场的活力。

Abstract:

Autonomous vehicle （AV） testing data is characterized by containing both continuous and discrete variables， and indirect information including timestamps， latitude and longitude， etc.， for which the traditional information entropy method should be adapted to AV testing data analyses. This paper proposes a driving mode-weighted information entropy method to determine the amount of AV testing data information by adapting the traditional entropy method using data feature mining and filtering， continuous variable discretization， driving mode-weighting， etc. Then， it establishes an AV testing data pricing model based on an evolved Stackelberg game with constraints of both information amount and data trading-platform profit rate， by integrating the information amount into the utility function of data consumers. It conducts a typical case analysis based on the actual test data of AV road testing in Shanghai. The results show that the evolved Stackelberg game-based pricing model can effectively evaluate the information amount of AV testing data， and reasonably allocate the profits among the three stakeholders of data providers， data trading platform， and data consumers. The data trading volume and the total utility of the system could be increased significantly， which contributes substantially to the booming of the AV testing data trading market.