In this paper, we present an improved comprehensive learning particle swarm optimiser (CLPSO) by using a generalised opposition-based learning concept (GOBL). The proposed approach, called GOCLPSO, employs similar schemes of opposition-based differential evolution (ODE) for opposition-based population initialisation and generation jumping with GOBL. Experimental studies on 13 benchmark functions show that GOCLPSO could achieve more accurate solutions than CLPSO for the majority of test cases.