During 802.11ax UL OFDMA magic happens in the RF space. What were RF collisions before is with 802.11ax normal behavior.

I will in this blog briefly explain what happens in the RF space during UL OFDMA.
I recommend reading one of the previous articles where I explain what it looks like in Wireshark.
I will simplify it so it is two clients that will send their data uplink in two different 106-tones RU

A short summary of the UL OFDMA process
– The AP sends a Basic Trigger frame, as a broadcast frame, in the legacy frame format at one of the mandatory rates
– STAs (clients) that are announced in the Basic Trigger frame sends their data uplink to the AP in parallel
— The AP sends a broadcast Multi-STA BlockAck frame to acknowledgment the received data frames

The Basic Trigger frame
This is the frame that is sent by the AP to solicits an uplink triggered frame from one or multiple STAs. This frame is a control type of frame and is sent at one of the mandatory data rates.
The information in the Common Info field and the User Info field is used by the STAs to build their uplink data frames
It uses the full channel, 20MHz. A visualization of this could be like this

 

The UL OFDMA data frame
The STAs (AID1 and AID2) will receive the Basic Trigger frame, interpret it and start sending their data in a frame format called Trigger Based frame (HE TB PPDU).

Each and every 802.11 frame has to start with the legacy preamble for backward compatibility, then these two have to send the pre-HE preamble because it is a 802.11ax frame. These two preambles are sent at the legacy subcarrier format, with 48 and 52 data subcarriers with 312.5KHz spacing. When this is done each STA will send the HE Short Training fields and the HE Long Training fields at the 802.11ax subcarriers format and at each STAs assigned subcarriers. And at last, each STA will send their data (A-MPDU) on their assigned subcarriers.
This transmission will start, from both STAs, a SIFS after receiving the Basic Trigger frame. A visualization of this is like this.

As we see, the legacy preamble and the pre-HE preamble uses the full channel and the HE-training fields and the A-MPDU is sent at each STAs assigned RU

If we combine these into one picture it looks like this, from the perspective of the AP

In legacy 802.11 (a,n and ac) this is a collision in the RF domain and the AP is not able to decode it correctly.
In 802.11 ax, the AP knows what is will receive and the legacy preamble of those two transmissions is equal because of the STAs (AID1 and AID2) uses information from the Basic Trigger frame to build the legacy preamble.
So the AP has basically one task: to use all its skills to prepare itself to receive a Multiple Input transmission (from MIMO). Spatial diversity, radio chains, memory, processor power, etc
And during the reception of the HE-training fields and the data (A-MPDU) the AP knows the setup (subcarriers, MCS, spatial streams, guard interval and so on).

And this works. My lab network based on Cisco equipment does UL OFDMA

 

The Multi-STA BlockAck
When the AP has received the data frame from the STAs, it acknowledgment the data frames with the Multi-STA BlockACK. This is a control frame sent at any of the basic rates, but it is usually sent at a mandatory data rate. Since it is sent at a basic rate it uses the full channel. Visualization like this

 

Summarize
This is UL ODFMA briefly explained