In　conventional　fuzzy　neural　network　based　on　sigmoid　function,　the　number　of　nodes　at　the　corresponding　intermediate　layer　grows　larger　with　the　increase　of　fuzzy　rule　numbers.　In　order　to　obtain　real　time　performance　of　the　control　system,　an　improved　fuzzy　Gaussian　radial　basis　function　neural　network　controller　for　omni-directional　intelligent　wheelchair　was　proposed.　The　structure　and　the　design　method　of　the　improved　controller　were　described　in　detail.　The　optimized　fuzzy　neural　network　reduced　the　numbers　of　intermediate　lay,　thus　the　computation　real　time　ability　could　be　guaranteed.　Furthermore,　the　stability　of　the　controller　was　proved　through　Lyapunov　stability　theory.　Finally,　the　proposed　optimized　controller　was　applied　to　an　omni-directional　intelligent　wheelchair,　and　the　simulation　results　of　tracking　control　verified　the　real　time,　effectiveness　and　feasibility　of　the　proposed　method.