To　achieve　a　robust　autopilot　for　ship　course-keeping,　a　controller　based　on　support　vector　machines　(SVM)　and　L2-gain　is　proposed.　A　cascaded　plant　consists　of　a　nonlinear　ship　response　model　and　the　rudder　actuator　dynamics.　Uncertainties　including　modeling　errors　and　external　disturbances　are　taken　into　account　in　the　response　model.　SVM　is　applied　to　identify　the　complicated　nonlinear　dynamics　including　the　modeling　errors,　while　L2-gain　design　is　used　to　suppress　the　uncertain　disturbances.　To　obtain　the　optimized　structural　parameters　in　SVM,　particle　swarm　optimization　(PSO)　method　is　incorporated.　Stability　analysis　and　simulation　results　demonstrate　the　validity　of　the　proposed　control　scheme.