direction. However, a single locomotion generator is not sufficient
to reach a distant goal in a complex environment. Similarly, a single
locomotion generator may become useless if the environment
changes, e.g. if the terrain changes from gravel to sand. In such
cases, the robot should be equipped with several locomotion generators,
which are switched by a high-level algorithm to achieve
the desired goal. For example, a robot equipped with ‘walk’ and
‘rotate-left’ actions can visit many places in an environment, but
these actions need to be combined to reach a goal and to avoid obstacles.
This can be achieved by using motion planning.
In this work, we present a novel high-level motion planning
technique for modular robots operating in complex environments.
The proposed planner finds a sequence of predefined motion
primitives to reach a desired goal. The motion primitives are the
vocabulary of the basic skills of the robot, such as ‘walk-forward’
and ‘rotate-left’. The robot can thus be equipped with several
primitives optimized for a given terrain, rather than having only a
single locomotion system that can be adapted. Such primitives can
be realized using simple methods, e.g. well-known CPGs, which can
be optimized for robots of arbitrary structure.
The text is organized as follows. Related work is summarized in
the next section. The hardware architecture of the robotic modules
developed in the Symbrion/Replicator projects is described in Section
3. A novel approach for global motion planning with motion
primitives is described in Section 4. The results of the simulations
are discussed in Section 5 and the results of the hardware experiments
are discussed in Section 6.