Integrated Motion Planning for a Hexapod
Robot Walking on Rough Terrain
Dominik Belter
∗
Piotr Skrzypczy´nski
∗
∗
Institute of Control and Information Engineering,
Poznan University of Technology
(e-mail: {Dominik.Belter, Piotr.Skrzypczynski}@put.poznan.pl).
Abstract: Missions of walking robots in distant areas require use of the teleoperation mode.
However, the capabilities of a human operator to sense the terrain and to control the robot
are limited. Thus, a walking robot should have enough autonomy to take an advantage of
its high locomotion capabilities in spite of a limited feedback from the remote operator. This
paper presents a method for real-time motion planning on a rugged terrain. The proposed
method employs several modules for planning the robot’s path and trajectories of the feet,
foothold selection, collision avoidance, and stability analysis. By using this method the robot
can autonomously find a path to the desired position and discriminate between traversable and
non-traversable areas. The rapidly exploring random trees concept is used as a backbone of the
proposed solution. Results of simulations and experiments on the real robot are presented.
Keywords: walking robot, path planning, map building, robot navigation, robust walking
1. INTRODUCTION
It is not possible for a remote operator to control a
walking robot by defining the movement of each leg. The
robot should have a goal which can be defined by the
operator, but gait generation and coordination of the legs
have to be solved on-board. When the robot is walking
on a rough, demanding terrain the task becomes much
harder, and the control system of the robot should provide
more autonomous decision capabilities. The main tasks
of the control system are: foothold selection, generation
of the feet trajectories, and global path planning for the
robot’s trunk. The obtained path should be executable,
which means that the kinematic constraints, the stability
maintenance criteria, and any possible collisions between
parts of the robot have to be taken into account.
The number of combinations of the possible robot’s pos-
tures and legs’ movements in a multi-legged robot makes
the motion planning task difficult and computationally
expensive. Therefore, some researchers try to simplify the
problem. Such a radical simplification can be achieved by
adapting existing 2D planning techniques to operate in 3D
terrain, and then by following the obtained path using a
reactive controller (Wooden et al. (2010)). Similarly, Bai
and Low (2001) mark obstacles on a legged robot’s path as
inaccessible or traversable and use a potential field algo-
rithm for path planning. These approaches may generate
quite conservative motion plans, which do not take full
advantage of the locomotion capabilities of a legged robot.
Another way of addressing the issue of computational
complexity is presented by Kolter et al. (2008), who at
first plan a path of the LittleDog robot’s trunk without
considering the trajectories of its feet. Then, the footholds
are planned to follow the initial path while considering
appropriate constraints. This approach restricts the huge
search space, but also may result in motion plans that do
Fig. 1. The Messor robot
not consider all possible paths, as some solutions could be
eliminated in an early stage of planning.
If more computing power is available, more complicated
motion planning strategies may be considered. A recent
work (Vernaza et al. (2009)) presents a search-based plan-
ning approach for a quadruped walking on rough terrain.
The motion planning problem is decomposed into two
main stages: an initial global planning stage, which results
in a footstep trajectory, and an execution stage, which
generates trajectories of the joints that enable the robot
to follow the footstep trajectory. Because the space of foot-
step trajectory is still high-dimensional the R
∗
algorithm,
which combines aspects of deterministic and randomized
search, is used for planning. Kalakrishnan et al. (2010)
apply a yet more extended hierarchy of planning and
optimization methods, which works well for a demand-
ing case of a quadruped robot walking dynamically on a
rough terrain. However, such an approach increases the
complexity of the whole control architecture. The high-
dimensionality issue in motion planning for legged robots
Preprints of the 18th IFAC World Congress
Milano (Italy) August 28 - September 2, 2011
Copyright by the
International Federation of Automatic Control (IFAC)
6918