In this paper, we propose an approach which ensures the dynamic stability of a biped robot called “BIPMAN”. It is based on the correction of the trunk center of mass acceleration and on the distribution of the forces exerted by the limbs on the trunk. This latter is performed by means of a linear programming method (the simplex method). The retained criterion allows to optimize force distribution as well as trunk roll and pitch angles. Weighting factors are introduced into this criterion in order to define criteria adapted to specific tasks. Modifying these factors is a solution to the task transition problem. Many simulation results are presented to demonstrate criteria and constraints influences on dynamic stability. They lead us to introduce a new approach called RTCA (Real Time Criteria and Constraints Adaptation). This relies on the analysis of position, velocity and acceleration vectors for criterion real time adaptation and on forces analysis for constraints real time adaptation. The RTCA approach is finally validated through simulations results for a specific task.