The problem of determining the optimal or suboptimal path in space is a
demanding task with multiple applicability as well as necessity for fast calculation.
The computational complexity increases depending on the increase in the number
of parameters or their values and the complexity of obstacles arrangement.
The master’s thesis is based on the use of RRT-methods with the fast presentation
feature of possible non-optimal movement through space in the form of a state
transition graph, within which the best path is determined by using the reverse
path tracking method. The acquired path quality is further improved by reducing
redundant turns and represented with Bezier polynomials in the continuous
smooth path form for the possibility of immediate tracking. Therefore with the
use of RRT-methods, a non-optimal state transition graph is rapidly yielded and
in combination with further processing, of improving the disadvantage of fast
construction, a high quality path outcome is thus quickly acquired. In the light
of speeding up the construction of RRT-trees, upgraded procedures of individual
steps are implemented, such as predefining the configuration space for faster tree
node collision checking with a possible nearby obstacle, the choice of using the
search for the nearest node only within the interesting part of the area in order to
reduce the number of checked nodes and taking into account the obstacle shape
when redefining a new node. The latter allows more efficient tree spreading along
an obstacle or narrow passages. Experiments confirms the effective use of simple
RRT methods, without subsequently linking of nearby states, in combination
with further path processing, which allows fast calculation of quality path.
|
