Scholarlink

<p>Significant safety, efficiency, and cost-effectiveness enhancements can be achieved by employing a deep neural network for vehicle automation. However, there are currently existing issues with applying deep networks to vehicle automation. The potential for algorithmic bias, which can result in discriminatory or unjust decisions, is one of the primary concerns. Another obstacle is the requirement for trustworthy and high-quality data. Deep network algorithms require large quantities of high-quality data to function properly, and obtaining this data in a vehicle control context can be challenging. This paper proposes a deep network optimized through intelligent computing to address these challenges and enhance intelligent vehicle driving. (1) An improved CSPLSTM model (ICSPLSTM) is developed for vehicle trajectory prediction by incorporating adaptive mechanisms. (2) The proposed model introduces an adaptive strategy that enables real-time adjustment of network parameters based on observed trajectory data, optimizing prediction in dynamic and complex real-world road environments. (3) The model employs online learning during inference, updating network weights through backpropagation to continuously improve predictive accuracy. Experiments indicate that the method described in this paper has a positive impact on the intelligent driving performance of vehicles.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A vehicle intelligent driving control method based on intelligent computing optimization deep network

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月12日

查看全文
  • Chaoming Ji

摘要

Significant safety, efficiency, and cost-effectiveness enhancements can be achieved by employing a deep neural network for vehicle automation. However, there are currently existing issues with applying deep networks to vehicle automation. The potential for algorithmic bias, which can result in discriminatory or unjust decisions, is one of the primary concerns. Another obstacle is the requirement for trustworthy and high-quality data. Deep network algorithms require large quantities of high-quality data to function properly, and obtaining this data in a vehicle control context can be challenging. This paper proposes a deep network optimized through intelligent computing to address these challenges and enhance intelligent vehicle driving. (1) An improved CSPLSTM model (ICSPLSTM) is developed for vehicle trajectory prediction by incorporating adaptive mechanisms. (2) The proposed model introduces an adaptive strategy that enables real-time adjustment of network parameters based on observed trajectory data, optimizing prediction in dynamic and complex real-world road environments. (3) The model employs online learning during inference, updating network weights through backpropagation to continuously improve predictive accuracy. Experiments indicate that the method described in this paper has a positive impact on the intelligent driving performance of vehicles.