Before we go into the details, let me clarify several issues about UDT's congestion control algorithm.
When we say "congestion control" here, we mean the mechanism to effectively utilize the network bandwidth. Another term, "flow control", is usually used to name the approach to prevent incoming traffic overwhelming the receiver. Congestion control is a more difficult problem.
Congestion control can tune the data sending rate in two ways: limit the number of on-flight packets and set the inter-packet time. TCP uses the window-based approach, whereas UDT uses the rate-based approach.
The most common congestion control algorithm is TCP's AIMD algorithm. We have already discussed AIMD.
Another important factor in congestion control is how to indicate a network congestion. Most protocols uses packet loss information, including both TCP and UDT. The increase in round trip time delay can also indicate congestion somewhere along the network.