In simple words: ResNet uses shortcuts to support training of deeper networks.
The difficulty
Adding more layers can make a network harder to train. Earlier information and learning signals may weaken as they pass through a very deep stack.
The shortcut
Residual Network (ResNet) uses skip connections. A block can pass its input forward and add a learned change to it. This gives information and gradients a more direct route through the network.
See the flow
input→learned block + shortcut→next block
Simple example
Instead of rewriting a document from scratch at each step, keep the original and add corrections. That is the intuition behind a residual connection.
Remember: ResNet uses shortcuts to support training of deeper networks.