Teleoperated robots offer a unique combination of robotic strength and precision, coupled with human judgment and ingenuity, to enhance productivity and safety in complex, unstructured, and hazardous industrial environments. However, teleoperation over ultra-remote distances has posed a significant challenge, with control performance being compromised by the inherent trade-off between distance and latency. This study presents a novel cloud-based human-robot motion mapping framework for ultra-remote teleoperation. An intuitive dual-arm human-robot motion mapping framework based on capturing human motion has been implemented, and private network connectivity is established using the Google Cloud platform for the efficient transfer of control data required for human-robot motion mapping. Moreover, a novel feedforward control framework is proposed to mitigate the negative impact of high network latency on teleoperation control performance. To demonstrate the feasibility of the proposed approach, an intercontinental teleoperation system between China and Sweden was developed, and a series of practical applications were demonstrated by capturing the motion of an operator in Sweden to control the dual-arm robot located over 7800 km away in China. The proposed feedforward control framework for intercontinental teleoperation has the potential utility for improving human-robot interaction performance in ultra-remote teleoperation scenarios with high network latency. More demonstration videos can be found at https://fsie-robotics.com/intercontinental-teleoperation/.