Around the world when peoples have independently invented the idea of farming. In each location they found something different to cultivate and in Africa one of the crops was pearl millet, Cenchrus americanus or Pennisetum glaucum. The earliest evidence was that this started around 4500 years ago, but the details have been debated.

Vavilov argued there was once source of domestication, while Harlan thought domestication occured across the Sahel, the strip of land south of the Sahara. It could be possible, as the wild progenitor of pearl millet is found across the Sahel.
Concetta Burgarella and colleagues have used whole-genome sequencing data to examine the domestication millet. They examined over 200 varieties to see how they differed. Usually the center of domestication is the region with the most genetic diversity. However, the team found that the diversity hotspots were in the western and eastern Sahel.
The genetic confusion seems to have been caused by cross-breeding domesticated millet with wild relatives, presumably because farmers were looking for some traits from those plants. They took this gene flow into account and then looked to see how the diversity differed across the current samples to find out where the millet spread from.
What they found was the millet of the Sahel probably didn’t come from the region we call the Sahel today. Writing in Nature Ecology and Evolution they say: “The spatial model identified an origin at latitude higher than the current range of wild populations in the central Sahel. The identified region corresponded to the Taoudeni Basin in the western Sahara (−6.61° E, 23.58° N), matching the wetter climate that prevailed in the Sahara 6,000 years BP. During that period, the Sahara hosted a rich Poaceae community and it was characterized by a widespread system of lakes. Drying of the Sahara led the plant communities to move south to their current distribution about 3,200 years ago.”
The archaeological evidence independently points to domestication in the same region. With two different methods coming to the same conclusion it seems that the debate between Vavilov and Harlan has been settled.