Figures and data in The quantitative and qualitative evaluation of a multi-agent microsimulation model for subway carriage design

Cite this article as: L. Cao, X. Li, F. Kang, C. Liu, F. Sun, R. Kotagiri; 2015; The quantitative and qualitative evaluation of a multi-agent microsimulation model for subway carriage design; International Journal of Microsimulation; 8(3); 6-40. doi: 10.34196/ijm.00120

17 figures and 3 tables

Figures

Figure 1

Download asset Open asset

The essential components and concepts of a typical multi-agent microsimulation model.

Figure 2

Download asset Open asset

The interior layout and exterior appearance of the SL Bombardier C20 subway carriage.

Figure 3

Download asset Open asset

The manual zone partition (No. 1 ∼ 19) of the focused part of an SL C20 carriage.

Figure 4

Download asset Open asset

The measurements of the simulated environment space E and the agent attributes p, q.

Figure 5

Download asset Open asset

An calculation example of dynamic comfort distance $q_{n}^{(t)}$ : auto-adaptation according to *N_t*.

Figure 6

Download asset Open asset

Illustration of agent turning angle that is selected obeying a discrete Gaussian distribution.

Figure 7

Download asset Open asset

The heterogeneous agent moving speeds are mapped to groups of Swedish population.

Figure 8

Download asset Open asset

The observation/simulation time (blue line) and the corresponding cubic interpolations (3D surface) as a function of existing and entering passengers/agents.

*Data source*: Table 1 and 2.

Figure 9

Download asset Open asset

The structure of a typical ELM network with one hidden layer (with L nodes), one input layer (with 23 nodes), and one output layer (with 10 nodes representing 10 categories).

Figure 10

Download asset Open asset

The evaluation results obtained via an unified multinomial classifier: (a) the performance sensitivity obtained via grid searching; (b) the class-wise confusion matrix of prediction accuracy.

(18)

Download asset Open asset

Figure 11

Download asset Open asset

The quantitative evaluation results using receiver operating characteristics (ROC) graph:

Figure 12

Download asset Open asset

A screen-shot of “left-right sub-carriage imbalance” in the SL C20 subway carriage.

Figure 13

Download asset Open asset

A screen-shot of “doorway jam” phenomenon in the SL C20 subway carriage.

Figure 14

Download asset Open asset

A new design of seat layout by emptying wall areas in the SL C20 subway carriage.

Figure 15

Download asset Open asset

An example of “pivot blocking” in the SL C20 carriage with the new seat distribution.

Figure 16

Download asset Open asset

The “left-right sub-carriage imbalance” problem caused by “vertical detour difficulty”.

Tables

Table 1

Real data collected from in-field observation.

d_i	Initial Status (entering\|existing passengers)	Real-data Dimensions (i.e. d_i,1, d_i,2,…, d_i,23)
d₁	17-01,19-03\|16:1,17:2,18:1,19:1	1,0,0,0,0,0,0,0,0,0,0,0,0,1,1,2,1,2,1,08,0,1,0
d₂	17-03,19-00\|14:1,16:3,17:2,18:2,19:4	0,1,0,0,0,0,0,0,0,0,0,0,0,2,1,1,2,4,4,12,4,3,0
d₃	16-01,18-03\|9:1,14:3,15:3,16:4,17:3,18:2,19:2	1,1,1,0,0,0,0,0,1,0,0,0,0,3,3,2,4,3,3,18,1,1,0
d₄	16-02,18-05\|1:1,14:3,15:3,16:4,17:2,18:4,19:3	3,2,3,0,0,0,0,0,0,0,0,0,0,4,4,3,3,2,3,20,3,1,1
d₅	16-07,18-05\|1:1,2:2,14:3,15:2,16:5,17:5,18:5,19:3	4,2,3,0,0,1,0,0,0,0,0,0,0,5,7,4,2,6,4,31,6,3,2
d₆	17-09,19-08\|1:2,2:4,14:2,15:2,16:7,17:3,18:4,19:4	3,3,2,0,0,1,1,0,0,0,1,0,0,6,7,6,5,4,6,28,3,1,1
d₇	16-10,18-06\|1:4,3:1,4:1,7:1,14:4,15:2,16:5,17:4,18:4,19:4	3,2,2,0,0,0,1,0,0,0,0,0,0,5,9,7,6,4,7,30,7,4,3
d₈	17-08,19-10\|1:2,2:2,3:2,4:1,8:1,14:2,15:2,16:5,17:2,18:6,19:5	4,4,3,1,0,0,0,1,0,0,0,0,0,5,8,3,8,7,4,35,0,3,5
d₉	17-11,19-13\|1:2,2:2,3:1,6:1,8:1,14:3,15:3,16:4,17:3,18:2,19:3	2,4,2,0,1,0,0,0,1,1,0,0,0,6,8,6,5,5,8,40,3,3,4
d₁₀	16-10,18-11\|1:3,2:4,3:2,6:1,7:1,10:1,14:5,15:4,16:5,17:4,18:4,19:5	4,6,3,0,0,1,1,1,0,1,0,0,0,8,9,8,7,5,6,46,4,6,5

Table 2

Partial simulation data samples generated from 100 valid micro-model executions.

${\hat{d}}_{i}$	Model Initialization (entering\|existing agents)	Features of Simulation Data $({\hat{d}}_{i, 1}, \dots, {\hat{d}}_{i, 23})$
${\hat{d}}_{1}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,4,0,3,1,05,00,1,1
${\hat{d}}_{2}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,1,1,0,0,0,0,0,0,0,0,0,1,0,1,1,0,3,10,00,0,4
${\hat{d}}_{3}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,1,0,0,0,0,0,0,0,0,0,0,1,0,3,1,1,1,13,00,2,3
${\hat{d}}_{4}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,3,1,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,18,04,4,4
${\hat{d}}_{5}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,2,1,0,0,0,0,0,0,0,0,0,1,0,1,1,1,1,10,02,3,5
${\hat{d}}_{6}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,3,1,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,09,03,3,4
${\hat{d}}_{7}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,3,1,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,10,02,2,4
${\hat{d}}_{8}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,1,1,0,0,0,0,0,0,0,0,0,0,0,1,1,1,3,08,01,1,3
${\hat{d}}_{9}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,3,1,3,1,05,00,1,2
${\hat{d}}_{10}$	17–01,19–03\|16:1,17:2,18:1,19:1	1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,4,0,1,3,07,00,1,1
${\hat{d}}_{20}$	17–03,19–00\|14:1,16:3,17:2,18:2,19:4	1,0,1,0,1,0,0,0,0,0,0,0,1,0,0,3,4,2,2,23,05,5,0
${\hat{d}}_{30}$	16–01,18–03\|9:1,14:3,15:3,16:4,17:3,18:2,19:2	1,1,1,0,1,0,0,0,0,0,0,0,1,1,2,5,1,4,4,27,02,3,1
${\hat{d}}_{40}$	16–02,18–05\|1:1,14:3,15:3,16:4,17:2,18:4,19:3	3,2,2,0,0,1,0,0,0,0,0,1,0,4,4,2,3,2,3,43,05,1,1
${\hat{d}}_{50}$	16–07,18–05\|1:1,2:2,14:3,15:2,16:5,17:5,18:5,19:3	3,3,2,0,0,1,0,0,1,0,1,0,1,5,5,4,3,6,3,46,07,2,2
${\hat{d}}_{60}$	17–09,19–08\|1:2,2:4,14:2,15:2,16:7,17:3,18:4,19:4	4,2,3,0,1,2,0,0,0,1,0,0,0,6,5,7,5,6,3,47,05,2,1
${\hat{d}}_{70}$	16–10,18–06\|1:4,3:1,4:1,7:1,14:4,15:2,16:5,17:4,18:4,19:4	3,1,2,1,0,1,0,1,0,1,0,0,0,4,7,8,6,5,6,52,10,3,4
${\hat{d}}_{80}$	17–08,19–10\|1:2,2:2,3:2,4:1,8:1,14:2,15:2,16:5,17:2,18:6,19:5	3,5,3,0,2,0,1,0,0,1,0,0,1,5,6,4,6,6,5,56,01,3,5
${\hat{d}}_{90}$	17–11,19–13\|1:2,2:2,3:1,6:1,8:1,14:3,15:3,16:4,17:3,18:2,19:3	2,5,3,0,0,1,1,0,1,0,0,1,1,7,5,6,5,5,6,63,04,2,4
${\hat{d}}_{100}$	16–10,18–111:3,2:4,3:2,6:1,7:1,10:1,14:5,15:4,16:5,17:4,18:4,19:5	5,6,2,0,2,1,1,0,1,0,1,0,1,7,7,8,6,6,6,69,04,5,5

Table 3

Parameters and measures of 10 one-vs.-rest binary classifiers (200 trials); the scores labeled with bold, asterisk, and underline represent the best, 2nd best, and the worst results respectively.

Parameters & Measures	Class. k=1	Class. k=2	Class. k=3	Class. k=4	Class. k=5	Class. k=6	Class. k=7	Class. k=8	Class. k=9	Class. k=10	Mean
Parameters & Measures	Lk	10³	10²	500	10³	2,000	200	200	10³	50	Mean	500	N/A
C_k (baseline)	10⁶	10³	10²	10³	10³	10²	10²	1	10	10²	N/A
Mean Acc.	0.879	0.916	0.897	0.996∗	0.991	0.906	0.992	0.998	0.872	0.996∗	0.944
Dev.	0.074	0.036	0.015∗	0.021	0.029	0.037	0.027	0.012	0.074	0.02	0.034
Precision	0.852	1	0.1	1	0.942	0.643	1	0.985∗	0.291	1	0.781
Recall	0.799	0.16	0.005	0.965	0.97∗	0.135	0.92	1	0.195	0.955	0.61

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Article PDF

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

BibTeX
RIS

Open citations (links to open the citations from this article in various online reference manager services)

Mendeley

The essential components and concepts of a typical multi-agent microsimulation model.

The interior layout and exterior appearance of the SL Bombardier C20 subway carriage.

The manual zone partition (No. 1 ∼ 19) of the focused part of an SL C20 carriage.

The measurements of the simulated environment space E and the agent attributes p, q.

An calculation example of dynamic comfort distance qn(t): auto-adaptation according to Nt.

Illustration of agent turning angle that is selected obeying a discrete Gaussian distribution.

The heterogeneous agent moving speeds are mapped to groups of Swedish population.

The observation/simulation time (blue line) and the corresponding cubic interpolations (3D surface) as a function of existing and entering passengers/agents.

The structure of a typical ELM network with one hidden layer (with L nodes), one input layer (with 23 nodes), and one output layer (with 10 nodes representing 10 categories).

The evaluation results obtained via an unified multinomial classifier: (a) the performance sensitivity obtained via grid searching; (b) the class-wise confusion matrix of prediction accuracy.

The quantitative evaluation results using receiver operating characteristics (ROC) graph:

A screen-shot of “left-right sub-carriage imbalance” in the SL C20 subway carriage.

A screen-shot of “doorway jam” phenomenon in the SL C20 subway carriage.

A new design of seat layout by emptying wall areas in the SL C20 subway carriage.

An example of “pivot blocking” in the SL C20 carriage with the new seat distribution.

The “left-right sub-carriage imbalance” problem caused by “vertical detour difficulty”.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

An calculation example of dynamic comfort distance $q_{n}^{(t)}$ : auto-adaptation according to N_t.