1: \begin{abstract}

2: Harmonic potentials provide globally convergent potential fields

3: that are provably free of local minima.

4: Due to its analytical format, it is particularly suitable

5: for generating safe and reliable robot navigation policies.

6: However, for complex environments that consist of a large number

7: of overlapping non-sphere obstacles,

8: the computation of associated transformation functions can be tedious.

9: This becomes more apparent when: (i) the workspace is

10: initially unknown and the underlying potential fields are updated constantly

11: as the robot explores it;

12: (ii) the high-level mission consists of sequential navigation tasks among numerous regions,

13: requiring the robot to switch between different potentials.

14: Thus, this work proposes an efficient and automated scheme to construct

15: harmonic potentials incrementally online as guided by the task automaton.

16: A novel two-layer harmonic tree (HT) structure is introduced

17: that facilitates the hybrid combination of oriented search algorithms for task planning

18: and harmonic-based navigation controllers for non-holonomic robots.

19: Both layers are adapted efficiently and jointly during online execution to

20: reflect the actual feasibility and cost of navigation within the updated workspace.

21: Global safety and convergence are ensured both for the high-level task plan

22: and the low-level robot trajectory.

23: Known issues such as oscillation or long-detours for purely potential-based methods

24: and sharp-turns or high computation complexity for purely search-based methods are prevented.

25: Extensive numerical simulation and hardware experiments

26: are conducted against several strong baselines.

27: \end{abstract}

28: