Hierarchical Architectures

Hierarchical architectures, also named deliberative control architectures, were used for many years since the first robots began to be built. Examples of such architectures and robots are SRI's Shakey [54], Stanford's CART [50], NASA's Nasrem system [42] and Isik's ISAM [32], among others. These architectures are based on a top-down philosophy, following a sense-plan-act model (see Figure 2.2). The control problem is decomposed into a set of modules, sequentially organized: first the perception module gets the sensory information, which is passed to the modeling module that updates an internal model of the environment; after that, planning is done using this internal model, and finally the execution module implements the solution with the appropriate commands for the actuators.

Figure 2.2: Sense-plan-act model

This model works very well when the environment in which the robot is working can be tailored to the task to be performed (e.g. industrial robots in factories, with magnetic beacons, marked paths, etc.). However, when the task is to be performed in an unknown, unpredictable, noisy environment, they fail to succeed, as the planning is usually out-of-date by the time it is being executed.

Another drawback of such architectures is their lack of robustness. Since the information is processed sequentially, a failure in any of the components causes a complete breakdown of the system.