Unequal Error Protection for Short Block Length Communications

Unequal Error Protection for Short Block Length Communications

Project Overview:

The current trend of increasing heterogeneity in communication networks brings a variety of communication needs and requirements, which often lead to different priorities. This is especially true in IoT applications which can generate messages with different degrees of importance. In this Project we introduce a priority-based coding scheme in the finite blocklength regime and apply the scheme to two different channels: the general discrete memoryless channel (DMC) and the AWGN channel. The scheme simultaneously encodes two messages, one with high and one with low priority, both requiring finite delay. The code structure allows the transmission of the high priority message with higher reliability and shorter decoding delay. We further derive tight and computationally efficient analytical upper bounds on the error probability in both the DMC and AWGN channel. 

Figure 1: Vehicle-to-Vehicle (V2V) Wireless Coordination


This approach has an immediate application, for example, in vehicle-to-vehicle (V2V) wireless coordination. More specifically, it applies to active road safety applications which are part of Vehicular Safety Communication in the US Federal and State Intelligent Traffic Systems projects. As a specific example, shown in the figure, suppose vehicle V1 wants to overtake vehicle V2 in a one-lane road while vehicle V3 is passing on the opposite side. In such a critical situation to avoid overtaking collision, a so called “overtaking vehicle warning” should be sent to V1 in a one hop connection from V2 (or V3) which informs V1 to stop the overtaking procedure. In this case, the high priority message is the “overtaking vehicle warning” message which has critical latency requirements and the low priority message might be a speed limit notification which is more delay tolerant and helps V1 to make an overtaking decision in the next attempt.

Figure 2: Coding Scheme

Coding Scheme:

The objective is to decode the high priority message in a shorter delay and with a lower error probability comparing to decoding the low priority message. The coding scheme should be such that, although the low priority message takes more symbols to be decoded, its error probability is higher than the error probability of decoding the high priority message after receiving less symbols. According to the requirements, we design a priority-based random encoding scheme based on partial superposition coding. The first part of the codewords are superposition encoding of the messages where the high priority message is considered to be the cloud center and the low priority message is the satellite message, whereas the rest of the codeword roughly just encode the low priority message.

Figure 3: Results for low and high probability message


Numerical results confirm better reliability and delay performance for the high priority message and also verify the tightness of the analytical upper bound even with very short blocklengths. In the figure the red plot is for the low priority message decoded with a longer delay (30 symbols) comparing to the blue plot which is for the high priority message decoded with a shorter delay (20 symbols). The green plot also refers to the high priority message but decoded with the same delay as the low priority message. As seen all requirements are satisfied; lower error probability and shorter delay for high priority message. Also it can be seen that the analytical upper bounds are tight specially when the SNR increases.