在外卖行业中，配送时效不仅影响着顾客满意度和平台的利润，还直接涉及对骑手的奖惩。因此，本研究从配送时效的角度，构建了考虑骑手奖惩的外卖配送多目标模型，并设计了改进 NSGA – II 算法进行求解。首先，该模型中利用软时间窗来构建线性顾客满意度函数，并结合基于配送时效的超时惩罚与绩效奖励，建立与距离相关的骑手收益目标，以及考虑订单抽成的平台利润目标。此外，设计了基于 KNN分类的改进 NSGA - II 算法进行优化。先利用基于欧式距离与时间窗的 KNN 分类生成初始路径；再采用结合前向连续交叉策略的改进 NSGA - II 算法进行优化输出帕累托解。最后，经过算例验证，本研究提出的算法展现出良好的性能表现，且经分析得出奖惩机制配置会对三方的效益产生不同影响。

Presently, the takeaway industry continues to demonstrate robust growth, with takeaway services becoming increasingly entrenched in people's daily lives. Nevertheless, given the differences focal points of platforms, consumers, and riders, a pronounced discord of interests has emerged among the three parties, even hindering the development of the industry. Distribution timeliness, a pivotal factor, impacts not only customer satisfaction and platform profits but also directly influences the rewards and punishments for riders. To address these challenges, this study proposes a multi-objective routing optimization problem for takeout delivery, considering the rewards and punishments for riders based on the perspective of distribution timeliness. The primary aim of this study is to further balance the interests of all parties dynamically. The study adopts a soft time window concept to portray customer expectations and tolerance of delivery time, and a linear time satisfaction function is constructed to assess customer satisfaction. Concurrently, the rider's rewards and penalties are designated, encompassing the rider's overtime penalty cost based on on-time rate, the order overtime penalty cost in step-pricing style, and the performance reward based on satisfaction. The functions of rider revenue, driving cost, waiting cost, and platform revenue are then combined to establish a complex objective optimization model for maximizing customer satisfaction, rider revenue, and platform profit. Secondly, this study proposes an enhanced NSGA-II algorithm based on KNN classification. The proposed algorithm first allocates orders based on geographic distance and time window, generates high-quality and diverse initial populations, and subsequently outputs a Pareto solution using the enhanced NSGA-II algorithm. To assess the efficacy of the proposed algorithm, this study utilizes two ordinary region examples and one commercial region example for testing purposes. It then compares this algorithm with the traditional NSGA-II and MOPSO algorithms for validation purposes. The convergence of the algorithm and the diversity of the solution set are further verified by analyzing the iterative convergence curves of each objective function as well as the indicators of spacing and maximum spread of the solution set. Finally, a comparative analysis of customer satisfaction, rider revenue, and platform profit under different reward and punishment rules is conducted. The study indicates that the platform's punishment measures can enhance delivery efficiency, foster customer satisfaction and augment platform revenue. Equitable sanctions have the potential to engender a mutually beneficial scenario for all three parties. While the incentive mechanism may temporarily reduce platform revenue, it can enhance delivery efficiency and rider revenue. Consequently, this study offers a theoretical foundation for the coordination of interests among multiple parties in the future of takeaway services and the establishment of a platform's reward and punishment mechanism for riders.