ALLIANCE is a software architecture that enables fault-tolerant cooperative control of heterogeneous mobile robots performing missions with loosely coupled, largely independent subtasks. The architecture allows robots to individually select appropriate actions based on mission requirements, other robots' activities, environmental conditions, and their internal states. ALLIANCE is a fully distributed, behavior-based system that uses mathematically modeled motivations (such as impatience and acquiescence) to achieve adaptive action selection. It ensures robust, reliable, and flexible responses to environmental changes and robot team modifications, such as mechanical failures, skill learning, or robot additions/removals. The architecture is demonstrated through an implementation on a team of robots performing a hazardous waste cleanup mission. The ALLIANCE architecture is designed to address the challenges of cooperative multi-robot systems, including problem formulation, communication, coherent action, and conflict resolution. It allows robots to adapt to dynamic environments and team changes, ensuring mission completion even with robot failures or environmental changes. The architecture is based on behavior-based control, where multiple task-achieving behaviors are active simultaneously, each receiving sensory input and controlling actuator output. Lower-level behaviors (competences) are suppressed by higher-level behaviors when necessary. The architecture includes several behavior sets, each corresponding to a high-level task-achieving function. Motivational behaviors control the activation of these behavior sets, with activation levels determined by factors such as sensory feedback, inter-robot communication, and internal motivations like impatience and acquiescence. The architecture incorporates parameters to manage robot impatience and acquiescence, allowing robots to adapt to task failures and environmental changes. The system ensures that robots can dynamically adjust their behavior sets based on task progress, team member actions, and environmental conditions. The architecture is tested in a hazardous waste cleanup mission, where robots must find spills, move them to a goal, and report progress. The results show that the architecture enables effective cooperation, even with robot failures or environmental changes. The system is designed to be fault-tolerant, reliable, and adaptable, ensuring mission completion in dynamic and unpredictable environments. The architecture is implemented on a team of robots, demonstrating its feasibility in real-world applications.ALLIANCE is a software architecture that enables fault-tolerant cooperative control of heterogeneous mobile robots performing missions with loosely coupled, largely independent subtasks. The architecture allows robots to individually select appropriate actions based on mission requirements, other robots' activities, environmental conditions, and their internal states. ALLIANCE is a fully distributed, behavior-based system that uses mathematically modeled motivations (such as impatience and acquiescence) to achieve adaptive action selection. It ensures robust, reliable, and flexible responses to environmental changes and robot team modifications, such as mechanical failures, skill learning, or robot additions/removals. The architecture is demonstrated through an implementation on a team of robots performing a hazardous waste cleanup mission. The ALLIANCE architecture is designed to address the challenges of cooperative multi-robot systems, including problem formulation, communication, coherent action, and conflict resolution. It allows robots to adapt to dynamic environments and team changes, ensuring mission completion even with robot failures or environmental changes. The architecture is based on behavior-based control, where multiple task-achieving behaviors are active simultaneously, each receiving sensory input and controlling actuator output. Lower-level behaviors (competences) are suppressed by higher-level behaviors when necessary. The architecture includes several behavior sets, each corresponding to a high-level task-achieving function. Motivational behaviors control the activation of these behavior sets, with activation levels determined by factors such as sensory feedback, inter-robot communication, and internal motivations like impatience and acquiescence. The architecture incorporates parameters to manage robot impatience and acquiescence, allowing robots to adapt to task failures and environmental changes. The system ensures that robots can dynamically adjust their behavior sets based on task progress, team member actions, and environmental conditions. The architecture is tested in a hazardous waste cleanup mission, where robots must find spills, move them to a goal, and report progress. The results show that the architecture enables effective cooperation, even with robot failures or environmental changes. The system is designed to be fault-tolerant, reliable, and adaptable, ensuring mission completion in dynamic and unpredictable environments. The architecture is implemented on a team of robots, demonstrating its feasibility in real-world applications.