Intelligent Perception and Control Technologies for Combine Harvesters in Complex Agricultural Environments: A Review

Combine harvesters in lodged, wet, weedy, uneven, or otherwise heterogeneous fields operate under rapidly changing feed rate, load, and material flow conditions. These disturbances often appear as drum overload, cleaning loss, grain breakage, impurity increase, and unstable travel, whereas conventional fixed-parameter operation still depends heavily on operator experience. This review examines intelligent perception and control technologies for combine harvesters from a mechanism-to-control perspective. The discussion covers dynamic load evolution, cleaning loss and grain damage mechanisms, multivariable coupling, pre-harvest perception, feed rate and internal state sensing, result layer loss and quality monitoring, forward speed control, threshing drum load regulation, adaptive cleaning control, and whole machine integration. The literature shows a clear shift from isolated sensing or single-parameter adjustment toward multimodal perception, state estimation, predictive control, digital twins, and edge deployment. At the same time, field robustness, cross-condition generalization, actuator bandwidth, sensing delay, and the coupling between result layer monitoring and closed-loop control remain the main barriers to deployment. The review, therefore, argues for a whole machine architecture that links environmental preview, internal state estimation, loss quality feedback, actuator-aware control, and cloud–edge–device collaboration for stable, low-loss, and autonomous harvesting in complex agricultural environments.