Page 327 - Introduction to AI Robotics
P. 327

310
                                                                                               Multi-agents
                                                                                            8
                                     on a single reactive agent, societies of reactive agents often exhibit an emer-
                                     gent societal behavior.
                                       Multi-agent societies can be characterized according to where they fall on
                                     at least four dimensions (since multi-agent theory is relatively new, even for
                                     robotics, new dimensions may surface over time). Heterogeneity refers to
                                     whether the member robots are identical in software and hardware. Cooper-
                                     ation may be either active or non-active, while control may fall in the spec-
                                     trum from fully distributed to fully centralized. A robot society may have
                                     a single, explicitly shared goal or each robot may have its own goal. When
                                     communication between agents is appropriate is a pervasive, open question.
                                       From a practical side, the emphasis in multi-agents has been on how fa-
                                     vorable group dynamics emerge from teams of homogeneous, purely reac-
                                     tive robots while operating under distributed control. Problems in emer-
                                     gent societal behaviors such as interference and the need to adapt to the
                                     open world can often been addressed by specifying social rules and internal
                                     motivation. However, more interest is emerging in robots that have either
                                     heterogeneous software or hardware capabilities, such as marsupial and re-
                                     configurable robots. The diversity of a heterogeneous team can be captured
                                     somewhat by the social entropy metric.


                               8.8   Exercises

                                     Exercise 8.1
                                     Give three reasons why multi-agents are desirable. Describe the general attributes of
                                     applications which are well-suited for multi-agents, and give one example.
                                     Exercise 8.2
                                     Define the following:
                                     a. heterogeneity
                                     b. control
                                     c. cooperation
                                     d. goals
                                     Exercise 8.3
                                     Consider the example of space ants. What would happen if the first robot commu-
                                     nicated with the other robots to recruit them to help move the asteroid? Would the
                                     behaviors or the goal structure necessarily change? Why or why not?

                                     Exercise 8.4
                                     Draw a FSA or write a script to coordinate the sequencing of the Pick Up the Trash
                                     behaviors for Io, Ganymede, and Callisto.
   322   323   324   325   326   327   328   329   330   331   332