One important requirement of modern supply chain management is the frequent exchange of containers via multiple cross-docks which requires spatial and time synchronisations between different types of vehicle. Moreover, as collaborations in logistics between several companies become popular, more flexible and extended models must be solved to consider the different needs of the companies. This is of high importance in a new logistics concept, the Physical Internet, which is expected to considerably improve the way logistics are handled in the current supply chain management.

To optimise the aforementioned requirements, a rich vehicle routing problem with pickup and delivery including numerous attributes is modelled and solved. A mathematical formulation is proposed and implemented in CPLEX to solve the problem. Given the complexity of the problem, solving large instances with exact methods is very time-consuming. Therefore, a multi-threaded meta-heuristic based on Simulated Annealing is developed. A set of new operators coupled with a restart strategy and memory are developed to help improve the performance.

Computational results on a generated data-set showed that the proposed meta-heuristic is superior to the CPLEX solver in terms of solvability and computational time. The proposed meta-heuristic was also compared with the best-known results by current state-of-the-art methods on a classical benchmark on pickup and delivery problems with time windows (with up to 200 customers). The experimental results showed that the proposed method was able to match the best-known results in many of these large scale instances

现代供应链管理的一项重要要求是通过多个交叉码头频繁更换集装箱，这需要不同类型车辆之间的空间和时间同步。此外，随着几家公司之间物流合作的流行，必须解决更灵活和扩展的模型以考虑公司的不同需求。这在新的物流概念“物理互联网”中具有非常重要的意义，它有望大大改善当前供应链管理中物流的处理方式。

为了优化上述要求，对包含众多属性的取件和配送的丰富车辆路径问题进行了建模和解决。提出了数学公式，并在CPLEX中实施以解决该问题。考虑到问题的复杂性，使用精确的方法解决大型实例非常耗时。因此，开发了一种基于模拟退火的多线程元启发式算法。开发了一组具有重新启动策略和内存的新运算符，以帮助提高性能。

对生成的数据集的计算结果表明，在可解性和计算时间方面，拟议的元启发式算法优于CPLEX求解器。在具有时间窗口（最多200个客户）的提货和交付问题的经典基准上，通过最新技术将提议的元启发式方法与最著名的结果进行了比较。实验结果表明，所提出的方法能够在许多大型实例中与最著名的结果相匹配。