The multileaf collimator (MLC) hardware constraints are usually neglected in the process of intensity-modulated beam optimization. Consequently, it is not always possible to deliver planned beam modulation using dynamic MLC. Beam optimization is significantly diminished if the results must be approximated due to limitations imposed by the delivery device. To overcome this problem, an inverse beam optimization method which incorporates the hardware constraints has been developed. The hardware constraints, including the leaf velocity, the dose rate and the minimum required gap between opposing and adjacent leaves, were considered. An iterative search for feasible modulation was conducted alternately in the dosimetric space and the MLC position-time space. The optimization algorithm was designed for a unidirectional leaf trajectory and a constant dose rate. A scheme to reduce tongue-and-groove underdosage during optimization was also implemented. Comparisons were made between the solutions produced by this method and conventional optimization disregarding the hardware restrictions. The beam profiles generated by the conventional method were modified to satisfy the hardware specifications. The results indicate that inclusion of MLC constraints during optimization can improve the degree of conformity that is deliverable.