The goal model used in our OMACS based systems has been formally defined and implemented in agentTool III. A runtime implementation used in our current multiagent and cooperative robotic projects has also been implemented. However, there are several ideas that have not been incorporated into GMoDS that could be very beneificial. These ideas include

1. The addition of maintenance goals (as opposed to achievement goals)
2. The addition of "simultaneous execution constraints" between goals 
3. The addition of the multiagent goals - requiring 'n' agents/robots to achieve a single goal

GMoDS currently only include goals of 'achievement'. That is, an agent pursues a set of actions that cause the goal to be achieved. Maintenance, on the other hand, may not require the agent to take any action, but only monitor the system to ensure that a specific condition is being maintained. Only when the condition is violated does agent need to take action. 

Simulatneous execution requires that two (or more) goals be pursued (assigned and achieved) at the same time. An example of two agents pursuing goals simultaneously would be one robot holding a door while a second robot carried an object through the door.

Requiring 'n' agents to work toward a common goal would allow us to model 'tightly coupled' goals where 'n' agents are required to work toward a single task. Generally, tightly coupled tasks also require simultaneous execution. An example would be multiple robots pushing a single box.

From a theoretical perspective, these poject need to address how they would be integrated into the current GMoDS specification and instance tree framework as well as their affect on goal achievement. Ideally, any extensions to GMoDS would be reflected in updates to agentTool, while changes in the execution semantics would be reflected in our GMoDS runtime implementation that is currently used on all MACR multiagent and cooperative robotics projects.