Edge computing does the processing on or near the device rather than sending everything to a central server. For a screen network the edge is the player, and it is usually where the interesting decisions belong.
Two reasons. Latency: an interaction that waits for a round trip feels broken, and a screen responding to a sensor cannot afford it. Resilience: a player that decides locally keeps working when the connection does not, which is the normal condition in retail, transport and anywhere temporary.
There is a privacy argument that is often the strongest one. Sensing that aggregates on the device and sends only counts never transmits an image at all, which converts a data-protection problem into a non-issue. If a proposal streams camera frames to a server for analysis, the same result is usually achievable at the edge with far less exposure.
The caveat is that edge devices are fleet infrastructure and get managed like appliances. Software at the edge needs versioning, monitoring and a remote update path, or you have a hundred slightly different deployments and no way to tell which is running what. The processing moves to the device; the operational burden does not disappear with it.