Exploration of unknown Environment (II.) exploration of autonomous environment based on Boundary point

(i) Exploration of the environment under the guidance of the guidance point

Through the analysis of the environment and the corresponding information processing, the paper extracts the target point of the guided robot to explore a series of unknown environment completely.

Explore the extraction of the guide point:

Acquiring the laser data--> extract the moving target point--> explore the decision Tree--> construct a complete or approximate complete environment map in a limited time.

A brief introduction to several methods: (strive to understand in general, best can be achieved)

1.A frontier-based approach for autonomous exploration

Detect open Space and unexplore space boundary point. (The problem is to move along the wall.) Before that, you can only explore a figure roughly by moving along the wall. or in an environment full of straight-line walls, corridors, round meeting rooms, etc.

The environment to solve: the exploration strategy under the complex environment. Based on grid maps. Evidence lattice method.

The key problem to explore and solve is: Given what you know about the world, where should you move to gain as much new information as possible?

Solution: To gain the most new information about the world, move to the boundary between open spaces and uncharted territory.

① Establish evidence lattice

open:occupancy probability (existence probability) <prior probability (prior probability)

Unknown: existence probability equals prior probability

Occupied: Existence probability is greater than prior probability

(My understanding is that a priori probability is 0.5. The present probability of an object obstacle is 1. The probability of existence is 0)

(a) evidence lattice (b) boundary (where obstructions are not in the obstruction of the moving target) (c) the smallest boundary area

Figure 0 and 1 correspond to open doors, 2 correspond to the end of the corridor that is not measured.

The open space is labeled as the frontier edge cell by the boundary of the unknown spaces, which can form a small area of the cells cluster. As long as the area is larger than the size of the robot (defined as the minimum required area), the region midpoint is considered to be the target point.

After the target point is found, the robot moves to the nearest point in the destination point that has not been visited.

Each time a target point is accessed, the point is added to the linked list that has been visited. and generate a new evidence lattice.

If the boundary target point cannot be reached after a period of time, add the point to the list of inaccessible frontiers. Again, 360 scans, heading down a little.

Experiment:

The robot is equipped with 1 laser rangefinder, 16 sonar sensors and 16 infrared sensors. The three sensors combine to construct evidence lattices, which are completely replaced by infrared sensors at very close distances (less than 16 inches).

Infrared and sonar are mainly used to avoid obstacles.

Below: white spaces with a lower occupancy rate, a small point of unknown occupancy, and a large black dot occupied.

(a) robot from the middle of the corridor, found 3 target points, toward the nearest 1 to (b) Update evidence lattice, found 2 target points, toward the near 0.

(e) 6 target points were found, but the last two were not able to go (at the table and Chair) (f) Half an hour map was built. All target points are detected and the rest are unreachable.

2. Search strategy based on RRT

The SRT method:randomized strategies for exploration

To be continue ...