Natural products, such as food products derived from animals and plants, exhibit a high degree of variability. Key factors for successful automation of these processes are the ability to sense workpiece variations and the flexibility to alter control strategies accordingly. This paper describes a robot system which incorporates complex sensing with knowledge based control. The system is used to remove the lateral pin bones from fish fillets by means of a robot manipulated waterjet cutter. A sensing system is used to obtain geometric and positional information about each fillet. A knowledge based control system uses a priori knowledge of the fish species and information from the sensing system to generate the appropriate robot commands. The successful deboning of fish fillets demonstrates the viability of this approach.>