1 Introduction
2 Literature review
2.1 Previous work on staggered car-following
2.2 Previous work on speed-LC relationship
2.3 Use of instrumented vehicles for traffic data extraction
3 Methodology for field data collection and analysis
3.1 Speed of test vehicle
3.2 LC between test and interacting vehicles
3.3 Speed of interacting vehicle
3.4 Determination of type of interacting vehicles
3.5 File-handling and processing of collected data
4 Field data collection
Serial no. | City | Road stretches | Test vehicles driven | ||||||
---|---|---|---|---|---|---|---|---|---|
Hatch-back car | Sedan car | SUV car | Van | Auto | Bike | Bus | |||
1 | Delhi | Outer ring road, Inner ring road, Nelson Mandela road, Africa Ave, GT road, Mehrauli-Badarpur road, Mehrauli-Gurgaon road, Charan Singh Road, and DND flyway | ✓ | ✓ | ✓ | ✓ | ✓ | ||
2 | Guwahati | Guwahati bypass, and GS road | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
3 | Kolkata | EM bypass, VIP road, CIT road, Gariahat road, DPS road (Tollygunge), Manicktala Vivekananda Road, Chittaranjan Ave, and Rash Behari Ave | ✓ | ✓ | ✓ | ✓ | ✓ | ||
4 | Bengaluru | Mysore Road, Bellary Road, CV Raman road, West chord road, Outer Ring Road, Hosur Road, NICE road, TCM Royan road (Majestic), and Racecourse road | ✓ | ✓ | ✓ | ✓ | ✓ | ||
5 | Pune | Bibwewadi Road, Sinhagad Road, Ambedkar-Wellesley road, Old Bombay-Poona Highway, Pune bypass, Ganeshkhind road, Tilak road, and Karwe road | ✓ | ✓ | ✓ | ✓ | ✓ | ||
6 | Mumbai | Link road, Western Express Highway, Eastern express highway, Eastern freeway, Pedder Road, Marine Drive, Bandra Worli Sea Link, and Jogeshwari-Vikhroli link road | ✓ | ✓ | ✓ | ✓ | ✓ |
5 Analysis and results
5.1 Relationship between lateral gap and average speed of interacting vehicle pairs
5.2 Evaluation of city-wise C versus \(\overline{\varvec{v}}\) data
City | Sample size | Intercept (All cities’ estimate: 120.883, SE: 2.232) | Slope (All cities’ estimate: 0.615, SE: 0.053) | ||||||
---|---|---|---|---|---|---|---|---|---|
Deviation | SE |
T
| Prob > |T| | Deviation | SE |
T
| Prob > |T| | ||
Bengaluru | 1310 | 0.123 | 3.854 | 0.032 | 0.975 | −0.025 | 0.099 | −0.250 | 0.803 |
Mumbai | 886 | −8.101 | 4.288 | −1.889 | 0.059 | 0.256 | 0.113 | 2.263 |
0.024
|
Pune | 918 | 9.671 | 4.971 | 1.945 | 0.052 | −0.351 | 0.119 | −2.959 |
0.003
|
Kolkata | 1174 | −2.565 | 3.757 | −0.683 | 0.495 | −0.188 | 0.097 | −1.933 | 0.053 |
Guwahati | 565 | 4.223 | 8.069 | 0.523 | 0.601 | 0.143 | 0.183 | 0.781 | 0.435 |
Delhi | 1163 | −3.351 | 3.51 | −0.955 | 0.340 | 0.164 | 0.077 | 2.134 |
0.033
|
5.3 Vehicle pairwise models for LC versus average speed
Pair | Intercept (I) | Slope (S) | ||||||
---|---|---|---|---|---|---|---|---|
Deviation | SE (σ) |
T
| Prob > |T| | Deviation | SE (σ) |
T
| Prob > |T| | |
Autos with other vehicles: I = 129.6 (σ = 5.2), S = 0.962 (σ = 0.2)
| ||||||||
Auto–Car | 7.555 | 6.67 | 1.133 | 0.258 | −0.516 | 0.242 | −2.128 |
0.034
|
Auto–Bike | −7.504 | 7.327 | −1.024 | 0.306 | 0.205 | 0.272 | 0.754 | 0.451 |
Auto–Auto | −0.872 | 9.337 | −0.093 | 0.926 | 0.171 | 0.372 | 0.461 | 0.645 |
Auto–HV | 0.821 | 11.791 | 0.07 | 0.945 | 0.139 | 0.457 | 0.305 | 0.760 |
Bikes with other vehicles: I = 119.76 (σ = 6.2), S = 0.541 (σ = 0.2)
| ||||||||
Car–Bike | 5.461 | 6.328 | 0.863 | 0.388 | 0.069 | 0.207 | 0.334 | 0.738 |
Auto–Bike | −3.728 | 6.668 | −0.559 | 0.576 | 0.248 | 0.218 | 1.14 | 0.255 |
Bike–Bike | −1.733 | 12.019 | −0.144 | 0.885 | −0.317 | 0.396 | −0.8 | 0.424 |
Cars with other vehicles: I = 126.51 (σ = 1.9), S = 0.550 (σ = 0.05)
| ||||||||
Car–Car | −9.125 | 2.392 | −3.415 |
0.001
| 0.128 | 0.064 | 1.994 |
0.046
|
Car–Bike | −1.297 | 2.655 | −0.488 | 0.625 | 0.061 | 0.076 | 0.801 | 0.423 |
Car–Auto | −2.482 | 3.024 | −0.821 | 0.412 | 0.096 | 0.085 | 1.124 | 0.261 |
Car–HV | −0.657 | 3.511 | −0.187 | 0.851 | 0.144 | 0.094 | 1.521 | 0.128 |
Car–LCV | 13.561 | 6.209 | 2.184 |
0.029
| −0.429 | 0.173 | −2.478 |
0.013
|
Vehicle pair | Sample size | Regression line | Coefficients of beta distribution | |||||
---|---|---|---|---|---|---|---|---|
Slope | Intercept |
α
1
|
α
2
|
a
|
b
|
p value of fit | ||
All combined | 6016 | 0.641 | 122.3 | 4.496 | 5.671 | −117 | 147.94 | 0.063 |
Auto–Auto | 140 | 1.037 | 114.55 | 1.861 | 2.338 | −89.79 | 112.73 | 0.787 |
Auto–Bike | 417 | 0.739 | 117.27 | 3.803 | 4.222 | −120.18 | 133.77 | 0.305 |
Auto–HV | 131 | 0.985 | 123.86 | 2.235 | 2.633 | −100.97 | 118.31 | 0.892 |
Bike–Bike | 87 | 0.224 | 118.02 | 1.511 | 1.856 | −68.93 | 84.55 | 0.782 |
Car–Auto | 897 | 0.642 | 124.12 | 4.263 | 4.995 | −120.59 | 141.41 | 0.600 |
Car–Bike | 1372 | 0.612 | 125.16 | 6.522 | 8.496 | −131.13 | 170.92 | 0.057 |
Car–Car | 2246 | 0.681 | 117.31 | 3.971 | 4.891 | −106.90 | 131.94 | 0.116 |
Car–HV | 538 | 0.675 | 126.35 | 2.264 | 2.789 | −88.39 | 108.85 | 0.997 |
Car–LCV | 139 | 0.132 | 139.78 | 1.711 | 2.213 | −74.98 | 96.64 | 0.861 |
5.4 Shying away behavior during interaction of heterogeneous vehicle pairs
5.5 Effect of carriageway width on LC versus average speed relationship
5.6 Study of LC in case of multiple vehicle interactions
-
Case 1: Interaction with vehicles is only on one side.
-
Case 2: Interaction is on both sides, and test vehicle is overtaking both the vehicles (V 1 > V 2, V 1 > V 3 in Fig. 11).
-
Case 3: Interaction is on both sides, and test vehicle is getting overtaken by other two vehicles. (V 1 < V 2, V 1 < V 3 in Fig. 11).
-
Case 4: Interaction is on both sides, and test vehicle is overtaking one vehicle and getting overtaken by other vehicle. (V 2 > V 1 > V 3).
Vehicle pair | Condition | Sample size | Regression line | Parameters of residuals beta distribution | |||||
---|---|---|---|---|---|---|---|---|---|
Slope | Intercept |
α
1
|
α
2
|
a
|
b
|
p-statistic | |||
Car–Car | Unconstrained | 1000 | 0.519 | 120.65 | 3.016 | 1.854 | −72.05 | 118.3 | 0.034 |
Constrained | 197 | 0.381 | 125.19 | 2.154 | 1.379 | −67.96 | 106.96 | 0.054 | |
Auto–Car | Unconstrained | 379 | 0.311 | 122.85 | 7.668 | 4.634 | −118.72 | 196.68 | 0.045 |
Constrained | 112 | 0.256 | 135.26 | 2.712 | 2.085 | −96.63 | 125.80 | 0.059 | |
Car–Bike | Unconstrained | 553 | 0.632 | 127.35 | 2.875 | 2.058 | −79.31 | 112.31 | 0.051 |
Constrained | 141 | 0.26 | 135.68 | 2.193 | 1.339 | −82.83 | 58.63 | 0.055 | |
Auto–Bike | Unconstrained | 151 | 0.864 | 126.46 | 1.276 | 1.322 | −87.57 | 86.59 | 0.075 |
Constrained | 67 | 0.171 | 136.60 | 2.063 | 1.868 | −91.84 | 103.12 | 0.071 |
5.7 Evaluation of model quality and comparison with earlier literature
Vehicle pair | Speed range (km/h) | ||||||
---|---|---|---|---|---|---|---|
0–10 | 10–20 | 20–30 | 30–40 | 40–50 | 50–60 | 60–70 | |
All combined | 0.294 | 0.338 | 0.187 | 0.052 |
0.045
| 0.362 | 0.874 |
Auto–Auto | N.D. | 0.762 | 0.886 | 0.178 | 0.355 | N.D. | N.D. |
Auto–Bike | 0.817 | 0.746 | 0.647 | 0.411 | 0.097 | N.D. | N.D. |
Auto–HV | 0.716 | 0.833 | 0.445 | 0.209 | 0.123 | 0.566 | N.D. |
Bike–Bike | N.D. | 0.710 | 0.667 | 0.314 | 0.789 | 0.077 | N.D. |
Car–Auto | 0.746 | 0.716 | 0.314 | 0.065 | 0.187 | 0.188 | N.D. |
Car–Bike | 0.098 | 0.184 | 0.174 | 0.188 |
0.031
| 0.657 | 0.435 |
Car–Car | 0.186 |
0.047
| 0.202 | 0.287 | 0.446 | 0.373 | 0.131 |
Car–HV | N.D. | 0.918 | 0.718 | 0.605 | 0.755 | 0.385 | N.D. |
Car–LCV | N.D. | N.D. | 0.863 | 0.164 | 0.373 | 0.547 | N.D. |
6 Conclusion
-
LC (C) versus average speed (\(\overline{v}\)) relationship of interacting vehicle pairs follows an upward linear trend with Beta-distributed residual; i.e., LC increase with an increase in interacting vehicles speed. Similar trend is observed in data collected from all six cities.
-
C versus \(\overline{v}\) relationships are modeled for various vehicle pairs such as car–car, car–bike and bike–auto. From the information of regression lines for models of various pairs, it is observed that motorized two-wheelers (bikes) maintain the least LC. Cars maintain the highest LC with light commercial vehicles (LCVs). Vehicles maintain lesser LC with heavy vehicles than with their own vehicle types due to their poor acceleration characteristics and maneuverability. For similar reasons, vehicles maintain higher clearances with autos.
-
LC maintained with the same vehicle type is lesser than that maintained with different vehicle types at similar speed levels. Thus, the LC between a pair of different vehicle types cannot be considered as the average of LC between the pairs of corresponding similar vehicle types.
-
It is observed that vehicles achieve maximum squeezing-in at a carriageway width of four lanes with paved shoulders.
-
When a vehicle interacts with multiple vehicles simultaneously, there is a compromise on LC at a particular average speed. The slope of LC versus average speed changes from 18% to 85% for different vehicle types in constrained versus unconstrained condition. However, intercepts remain consistent, indicating compromise in safety at higher speed ranges only.