T2b-dtspn - Data collection path planning with curvature-constrained trajectory - Dubins TSPN

The main task is to implement two approaches solving the curvature-constrained Traveling Salesman Problem with neighborhoods.

Due date December 01, 2024, 23:59 PST
Deadline January 11, 2025, 23:59 PST
Points 7
Label in BRUTE t2b-dtspn
Files to submit archive with DTSPNSolver.py
Resources B4M36UIR_t2_resource_pack
For Ubuntu 22.04 and newer pydubins
LKH


Necessary information in Lectures

Decoupled Approach

In the plan_tour_decoupled function implement the decoupled solution for the Dubins TSP with Neighborhoods (DTSPN) with disk-shaped regions.

The decoupled approach comprises the following basic steps

  1. Estimate sequence of visits by Euclidean TSP connecting centers of the regions.
  2. For each region, sample boundary points and heading angles.
  3. Find the shortest feasible tour comprising Dubins maneuvers connecting the regions, where the sequence of visits is estimated from the ETSP.

The plan_tour_decoupled function has the following prescription:

def plan_tour_decoupled(goals, sensing_radius, turning_radius):
    """
    Compute a DTSPN tour using the decoupled approach.  
 
    Parameters
    ----------
    goals: list Vector3
        list of the TSP goal coordinates (x, y, 0)
    sensing_radius: float
        neighborhood of TSP goals  
    turning_radius: float
        turning radius for the Dubins vehicle model  
 
    Returns
    -------
    Path, path_length (float)
        tour as a list of robot configurations in SE3 densely sampled
    """


Noon-Bean Transform Approach

In the plan_tour_noon_bean function implement the Noon-Bean trasnform solution for the Dubins TSP with Neighborhoods (DTSPN) with disk-shaped regions.

The Noon-Bean approach comprises the following basic steps

  1. For each region, sample boundary points and heading angles.
  2. Construct a distance matrix using the Noon-Bean transform (from lectures) where the individual regions correspond to the NoonBean's Generalized TSP sets.
  3. Find the shortest feasible tour created from Dubins maneuvers by solving the Asymmetric TSP problem characterized by the distance matrix.

The plan_tour_noon_bean function has the following prescription

def plan_tour_noon_bean(goals, sensing_radius, turning_radius):
    """
    Compute a DTSPN tour using the NoonBean approach.  
 
    Parameters
    ----------
    goals: list Vector3
        list of the TSP goal coordinates (x, y, 0)
    sensing_radius: float
        neighborhood of TSP goals  
    turning_radius: float
        turning radius for the Dubins vehicle model  
 
    Returns
    -------
    Path, path_length (float)
        tour as a list of robot configurations in SE3 densely sampled
    """

The LKH solver seems to be very sensitive to the choice of bigM constant for Noon-Bean transformation. It influences the computational time significantly. The lower the bigM constant is, the faster it computes; but if it is too small, it can violate the original assumption to visit each set only once. The reference sets bigM to be the length of the longest Dubins path and use it also for representing infinity. The distance matrix is normalized before solving by LKH, and thus it cannot contains infinity values. (Alternatively, you can use a slightly smaller number of samples.)


Appendix


Installation of the Prepared Dependencies

1) Download prepared codes and configuration files.

2) Install dubins package to python3.

On Ubuntu 18.04 or Ubuntu 20.04 (or Python < 3.9), install the package using

pip3 install dubins # or our favorite way to install the package
On Ubuntu 22.04 (or Python >= 3.9), the package is not available in pip, and it needs to be installed from the provided source (pydubins in b4m36uir_23_dubins_resource_pack.zip) located inside your resource pack directory using
./install_dubins.sh

3) The LKH solver (implementation of the Lin–Kernighan heuristic algorithm) and the GDIP Dubins library are already part of the resource package and they are installed as follows

./install_lkh.sh

Since both the dubins package and the LKH wrapper used the second part of the task use SE(2) coordinates represented as tuples, pose_to_se2 are se2_to_pose functions are provided for you convenience:

def pose_to_se2(pose):
    return pose.position.x,pose.position.y,pose.orientation.to_Euler()[0]
 
def se2_to_pose(se2):
    pose = Pose()
    pose.position.x = se2[0]
    pose.position.y = se2[1]
    pose.orientation.from_Euler(se2[2],0,0) 
    return pose


Dubins maneuver

courses/uir/hw/t2b-dtspn.txt · Last modified: 2024/09/16 20:14 by pragrmi1