Academia.eduAcademia.edu

Figure 4 – uploaded by Michael Lindenbaum

See full PDFdownloadDownload figure
Covering by ANT-WALK-1, shape #3, on a 14 x 14 matrix Num of Ants = 3; Total area = 161; Cleaned area = 161 Fig. 4. Simulated evolution of three agents, oriented by the ANT-WALK-1 rule. The gray-level is proportional to the number of visits in each tile. Note that most (90%) of the area is cleaned within 80% of the time.

Figure 4 Covering by ANT-WALK-1, shape #3, on a 14 x 14 matrix Num of Ants = 3; Total area = 161; Cleaned area = 161 Fig. 4. Simulated evolution of three agents, oriented by the ANT-WALK-1 rule. The gray-level is proportional to the number of visits in each tile. Note that most (90%) of the area is cleaned within 80% of the time.

Related Figures (14)

Fig. 1. System of rooms divided into square tiles with a dynamic obstacle. Two cleaning robots are shown with their directional smell traces. Note that the traces degrade with time.
Fig. 1. System of rooms divided into square tiles with a dynamic obstacle. Two cleaning robots are shown with their directional smell traces. Note that the traces degrade with time.
Fig. 2. Four cycles OF ANT-WALK-1 are needed to traverse a floor with four tiles. Also shown is the corresponding directed graph with the smell-labels on its edges. Note that there are two labels on each edge, to designate the trace intensity in each direction of the edge.
Fig. 2. Four cycles OF ANT-WALK-1 are needed to traverse a floor with four tiles. Also shown is the corresponding directed graph with the smell-labels on its edges. Note that there are two labels on each edge, to designate the trace intensity in each direction of the edge.
Fig. 3. Hard case for the ANT-WALK-1 rule. There are n vertices, and about 1.25n edges. The diameter is about 0.8, and the time needed to traverse it may be as long as O(n"). The dotted arrows show the worst case where each triangle of vertices is a “trap” that causes the ant to go back to its starting point.
Fig. 3. Hard case for the ANT-WALK-1 rule. There are n vertices, and about 1.25n edges. The diameter is about 0.8, and the time needed to traverse it may be as long as O(n"). The dotted arrows show the worst case where each triangle of vertices is a “trap” that causes the ant to go back to its starting point.
Covering by ANT-WALK-1, shape #3, on a 14 x 14 matrix
Covering by ANT-WALK-1, shape #3, on a 14 x 14 matrix
Covering by ANT-WALK-2, shape #3, on a 14 x 14 matrix
Covering by ANT-WALK-2, shape #3, on a 14 x 14 matrix
Covering by ANT-WALK-2, shape #3, on a 14 x 14 matrix
Covering by ANT-WALK-2, shape #3, on a 14 x 14 matrix
Fig. 8. Time of covering t;, versus k—the number of robots, with varying amount of sensing-noise, for the three ANT-WALK algorithms. Note the different scales on the vertical axis. It can be seen that ANT-WALK-2 has best performance in the given test case. Fig. 7. Three agents, oriented by the ANT-WALK-2 rule, with sensory-noise level of ten units. Note that even with noise, this algorithm is much faster than ANT-WALK-1; it is also advantageous over the ordinary DFS which is extremely vulnerable to sensing errors.
Fig. 8. Time of covering t;, versus k—the number of robots, with varying amount of sensing-noise, for the three ANT-WALK algorithms. Note the different scales on the vertical axis. It can be seen that ANT-WALK-2 has best performance in the given test case. Fig. 7. Three agents, oriented by the ANT-WALK-2 rule, with sensory-noise level of ten units. Note that even with noise, this algorithm is much faster than ANT-WALK-1; it is also advantageous over the ordinary DFS which is extremely vulnerable to sensing errors.
Fig. 9. The same data of the previous figure is here sorted by noise-level. It clearly shows the VERTEX-ANT-WALK is best in the absence of noise, while ANT-WALK-2 wins in noisy conditions.
Fig. 9. The same data of the previous figure is here sorted by noise-level. It clearly shows the VERTEX-ANT-WALK is best in the absence of noise, while ANT-WALK-2 wins in noisy conditions.
Fig. 10. Time of covering t;, versus k—the number of robots, with varying amount of sensing-noise, for the three ANT-WALK algorithms, averaging on 10 coverings.
Fig. 10. Time of covering t;, versus k—the number of robots, with varying amount of sensing-noise, for the three ANT-WALK algorithms, averaging on 10 coverings.
Fig. 11. Same data (average of 10 covers) of the previous figure, sorted by noise-level. Here again ANT-WALK-2 wins with noisy sensors.
Fig. 11. Same data (average of 10 covers) of the previous figure, sorted by noise-level. Here again ANT-WALK-2 wins with noisy sensors.
Covering by ANT-WALK, shape #5, on a 30 x 30 matrix Num of Ants = 10; Total area = 338; Cleaned area =
Covering by ANT-WALK, shape #5, on a 30 x 30 matrix Num of Ants = 10; Total area = 338; Cleaned area =
Covering by VERTEX~ANT-WALK, shape #0, on a 8 x 8 matrix
Covering by VERTEX~ANT-WALK, shape #0, on a 8 x 8 matrix
Covering by VERTEX-ANT-WALK, shape #0, on a 8 x» 8 matrix
Covering by VERTEX-ANT-WALK, shape #0, on a 8 x» 8 matrix
Covering by VERTEX-ANT-WALK, shape #0, on a 8 x 8 matrix
Covering by VERTEX-ANT-WALK, shape #0, on a 8 x 8 matrix
Related topics:
Mechanical EngineeringMulti Agent SystemRobot kinematicsManufacturing EngineeringShared memoryAdaptive SystemElectrical and Electronic Engineering
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved