Plant breeding programs evolve by developing a better understanding of genetic architecture of agricultural traits and making more use of it with minimal resource diversion from mainstream genetic gain accumulation. Thanks to new capacities for monitoring genome segregation, analytical approaches have evolved, shifting from specific experiments with ad hoc progenies to global analyses incorporated along the improvement process through new types of populations. An array of such populations is described here in order to illustrate current efforts at Cirad and IRD on a range of tropical crops. The leaders of these efforts and contributors of presentation materials are indicated in parentheses. Annual crops allow quick constitution and management of novel populations which have specific features: - Nested-Association-Mapping (NAM) populations in rice developed at CIAT and AfricaRice with IRD (Mathias Lorieux) - Back-cross NAM (BCNAM) populations developed in sorghum at IER with CIRAD and ICRISAT (Jean- François Rami) - Chromosome-segment substitution lines (CSSL) in groundnut developed at ISRA - CERAAS with Cirad, EMBRAPA and ICRISAT (Daniel Fonceka) - Recurrent Selection populations in rice developed and used at CIAT with CIRAD for genomic selection (Cécile Grenier). Vegetatively propagated fruit crops often find value in seedless types and often feature interspecific and intergenomic hybridization that fixed heterozygosity and sterility: - banana and plantain improvement requires pre-breeding prior to sterility-generating combinations, leading attention to the causes for restriction to recombination and segregation distorsions (Angélique D'Hont). Perennial crops often have a peculiar history that heavily impacted their population structure: - sugarcane has a short breeding history featuring interspecific hybridization from a limited number of founders which led to strong linkage disequilibrium (LD) and allows genome-wide association studies - cacao has a history of mi