In　traditional　method　to　avoid　obstacles　for　intelligent　wheelchairs,　the　fuzzy　logic　based-design　of　parameters　depends　on　the　designer＇　experiences.　Thus,　on　the　basis　of　fuzzy　neural　networks,　a　self-learning　obstacle　avoidance　algorithm　of　intelligent　wheelchairs　was　proposed.　The　algorithm　combined　fuzzy　logic　and　neural　networks　with　their　respective　advantages,　and　state　control　variables　were　used　to　record　omni-wheelchair　state　of　motion　to　solve　the　selection　problem　of　the　user　expecting　target　direction　and　wheelchair　obstacle　avoidance　direction.　Obstacle　avoidance　path　was　optimized　to　better　meet　the　needs　of　the　users　in　the　comfort　and　security　of　the　intelligent　wheelchair.　Simulation　and　physical　experiments　show　that　the　algorithm　improves　the　intelligence　of　obstacle　avoidance　and　comfort　of　the　wheelchair　and　can　be　used　in　the　wheelchair　obstacle　avoidance　controls.