Self-adaptive pervasive systems often implement adaptation in a centralised manner, where one component holds all the necessary knowledge to identify when and how the system needs to adapt. In self-adaptive pervasive systems, composed of autonomous components with different authorities (such as security, distribution, etc.), this approach cannot be implemented as composing a centralised knowledge is not feasible and it also obstructs the system's ability to dynamically change its components. A simple alternative would be to allow each component to adapt independently but this can quickly give rise to conflicts, race conditions and oscillations between multiple independent adaptations. To avoid these problems, we propose to coordinate individual adaptations so that each component's adaptation goals are satisfied. Each component proposes an adaptation which is reviewed by other components who may propose their own adaptations that they may need to do. This continues until a complete adaptation plan is agreed upon. In cases where certain individual adaptations conflict with some components' goals, components are instructed to seek alternative proposals. The Adaptation Manager component is in charge of the negotiation process and it also has the authority to resolve certain conflicts between adaptations. Our approach is evaluated in the context of pervasive workflow systems where the failure probability and execution times are assessed.