This is verified in the Steps and Outdoor-like scenarios. As was
shown in the Steps experiment, the robot is not able to overcome
the steps if no ‘raise-head’ primitive is integrated into the RRTMP
algorithm. Similar results are obtained on the Surface scenario
with the S-bot robot, where two types of motion primitives are
used: primitives learned on the plane and primitives learned on
the surface. As the size of the S-bot robot is comparable with the
size of the hills and dips on the surface, it climbs down into a dip in
certain situations. As the primitives learned on the plane are used,
the robot is not able to climb up from the dip and the success ratio
of the planning algorithm decreases. This situation occurs even if
the surface-learned primitives are used, as these reflect only on a
part of the surface and are not ready to cope with the dips. We can
conclude, that the motion primitives need to be prepared carefully,
taking into account the environment that the robot operates
in.
This is verified in the Steps and Outdoor-like scenarios. As wasshown in the Steps experiment, the robot is not able to overcomethe steps if no ‘raise-head’ primitive is integrated into the RRTMPalgorithm. Similar results are obtained on the Surface scenariowith the S-bot robot, where two types of motion primitives areused: primitives learned on the plane and primitives learned onthe surface. As the size of the S-bot robot is comparable with thesize of the hills and dips on the surface, it climbs down into a dip incertain situations. As the primitives learned on the plane are used,the robot is not able to climb up from the dip and the success ratioof the planning algorithm decreases. This situation occurs even ifthe surface-learned primitives are used, as these reflect only on apart of the surface and are not ready to cope with the dips. We canconclude, that the motion primitives need to be prepared carefully,taking into account the environment that the robot operatesin.
การแปล กรุณารอสักครู่..
