In this work, the problem of designing sensor-based controllers allowing to navigate in orchards is considered. The navigation techniques classically used in the literature rely on path following using metric maps and metric localization obtained from onboard sensors. However, it appears promising to use sensor-based approaches together with topological maps for two main reasons: first, the environment nature is rather changing and second, only high-level information are sufficient to describe it. One of the key maneuver when navigating through an orchard is the u-turn which must be performed at the end of each row to reach the next one. This maneuver is generally performed using only dead reckoning because of the lack of dedicated sensory data. In this paper, we propose two sensor-based control laws allowing to perform u-turns, improving the performance quality. They allow following particular spirals which are defined from laser rangefinder data and adapted to realize the desired maneuver. Their stability is studied and their performances are thoroughly examined. Finally, they are embedded in a complete navigation strategy to show their efficiency in our agricultural context