A method of beam failure recovery for a serving cell is disclosed. In response to a beam failure on a first bandwidth part in a serving cell, a first device detects a first candidate beam associated with the first bandwidth part (310). In response to a failure of detecting the first candidate beam, the first device switches from the first bandwidth part to a second bandwidth part in the serving cell. The second bandwidth part is configured with at least one reference signal identifying a second candidate beam for recovery from the beam failure (320). The first device detects the second candidate beam based on the at least one reference signal (330). In response to a success of detecting the second candidate beam, the first device transmits information concerning the second candidate beam to a second device for the recovery from the beam failure (340).