TY - JOUR
T1 - Stack-RRT*
T2 - A Random Tree Expansion Algorithm for Smooth Path Planning
AU - Liao, Bin
AU - Hua, Yi
AU - Wan, Fangyi
AU - Zhu, Shenrui
AU - Zong, Yipeng
AU - Qing, Xinlin
N1 - Publisher Copyright:
© 2023, ICROS, KIEE and Springer.
PY - 2023/3
Y1 - 2023/3
N2 - Most RRT-based extension algorithms can generate safe and smooth paths by combining parameter curve-based smoothing schemes. For example, the Spline-based Rapidly-exploring Random Tree (SRRT) guarantees that the generated paths are G2-continuous by considering a Bezier curve-based smoothing scheme. In this paper, we propose Stack-RRT*, a random tree expansion method that can be combined with different parameter curve-based smoothing schemes to produce feasible paths with different continuities for non-holonomic robots. Stack-RRT* expands the search for possible parent vertices by considering not only the set of vertices contained in the tree, as in the RRT-based algorithm, but also some newly created nodes close to obstacles, resulting in a shorter initial path than other RRT-based algorithms. In addition, the Stack-RRT* algorithm can achieve convergence by locally optimizing the connection relation of random tree vertices after each expansion. Rigorous simulations and analysis demonstrate that this new approach outperforms several existing extension schemes, especially in terms of the length of the planned paths.
AB - Most RRT-based extension algorithms can generate safe and smooth paths by combining parameter curve-based smoothing schemes. For example, the Spline-based Rapidly-exploring Random Tree (SRRT) guarantees that the generated paths are G2-continuous by considering a Bezier curve-based smoothing scheme. In this paper, we propose Stack-RRT*, a random tree expansion method that can be combined with different parameter curve-based smoothing schemes to produce feasible paths with different continuities for non-holonomic robots. Stack-RRT* expands the search for possible parent vertices by considering not only the set of vertices contained in the tree, as in the RRT-based algorithm, but also some newly created nodes close to obstacles, resulting in a shorter initial path than other RRT-based algorithms. In addition, the Stack-RRT* algorithm can achieve convergence by locally optimizing the connection relation of random tree vertices after each expansion. Rigorous simulations and analysis demonstrate that this new approach outperforms several existing extension schemes, especially in terms of the length of the planned paths.
KW - Continuous curvature
KW - non-holonomic robots
KW - path planning
KW - rapidly-exploring random tree (RRT)
KW - smooth path planning
UR - http://www.scopus.com/inward/record.url?scp=85147759452&partnerID=8YFLogxK
U2 - 10.1007/s12555-021-0440-2
DO - 10.1007/s12555-021-0440-2
M3 - 文章
AN - SCOPUS:85147759452
SN - 1598-6446
VL - 21
SP - 993
EP - 1004
JO - International Journal of Control, Automation and Systems
JF - International Journal of Control, Automation and Systems
IS - 3
ER -