Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/309322260

Predicting Optimal Trajectories for Constrained Pedestrian Turning Manoeuvres

Conference Paper · October 2016

CITATIONS READS
2 170

3 authors, including:

Charitha Dias Miho Iryo-Asano


Qatar University Nagoya University
47 PUBLICATIONS   210 CITATIONS    61 PUBLICATIONS   564 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Pedestrian Behaviors and Control Policy in NYC View project

All content following this page was uploaded by Charitha Dias on 21 October 2016.

The user has requested enhancement of the downloaded file.


Proceedings of the 8th International Conference on Pedestrian and Evacuation Dynamics (PED2016)
Hefei, China – Oct 17 – 21, 2016
Paper No. 15

Predicting Optimal Trajectories for Constrained Pedestrian


Turning Manoeuvres
Charitha Dias1, Majid Sarvi2, Miho Iryo-Asano3
1,3
Institute of Industrial Science, the University of Tokyo
Meguro-ku, Tokyo 153-8505, Japan
cdias@iis.u-tokyo.ac.jp; m-iryo@iis.u-tokyo.ac.jp
2
Department of Infrastructure Engineering, the University of Melbourne
Melbourne, Victoria 3010, Australia
msarvi@unimelb.edu.au

Abstract - Reproducing realistic walking trajectories associated with complex geometries and floor layouts, such as
turning configurations, in microscopic simulation models is beneficial in many applications. Estimating reliable
evacuation times for complex scenarios, 3-D visualization of realistic microscopic movements and virtual reality
applications (for example, in driving simulators) are few of those important applications. However, it is uncertain that
existing approaches capture the complete picture of turning manoeuvres with regards to walking paths, speed profiles
and acceleration-deceleration patterns. In this paper, minimum-jerk theory (‘jerk’ is defined as the time derivative of
acceleration) is utilized to model trajectories through different turning configurations. Normal speed walking (average
free-flow speed ≈ 1.4 m s-1) through 90°, 135° and 180° turning configurations are considered in this paper. Modelled
trajectories are compared with trajectories collected through controlled experiments qualitatively and quantitatively.
This comparison reveals a favourable match between modelled and experimental trajectories. This indicates that the
minimum jerk theory can be utilized to model trajectories through turning configurations under different constrained
boundary conditions. Output of this study could be beneficial not only for validating existing microscopic simulation
models but also for building new simulations based on more realistic representation of pedestrians’ movement
mechanisms through complex geometrical settings.

Keywords: Turning movements, Minimum-jerk theory, Human movement mechanisms, Optimal


trajectories, Pedestrian behaviour modelling

1. Introduction
Currently, microscopic pedestrian simulation tools are widely being utilized in practice such as in
optimizing public building designs [1]. However, reliability and precision of those software tools, which
are based on well-known microscopic models, are still questionable [2-3]. In order to enhance the reliability
of predictions by a model and for model calibration and validation purposes reliable empirical data are
required. Such empirical data should be collected under a variety of situations. That is because, microscopic
pedestrian behavioural patterns might significantly differ from one situation to another particularly in case
of complex scenarios. Understanding pedestrian movement mechanisms associated with complex settings
is important before modelling and simulating pedestrian movements realistically.
In the context of modelling microscopic pedestrian dynamics associated with rounding corners, several
attempts can be found in the literature. The simplest approach is to locate one or several intermediate points
to guide the desired movements around corners [4-5]. Desired path of a simulated pedestrian is planned
through these intermediate destinations. Such simplified approaches may reproduce unrealistic trajectories
and speed patterns as demonstrated in [6]. Instead of guiding points, Chraibi et al. [7] modelled desired
direction using guiding lines combined with update rules. These update rules have basically considered the
occupancy of guiding lines. Heuristic based methods [8-9] and rule based methods [10-11] have also been
implemented by several researchers. Various behavioural phenomena, for ex., pedestrians’ desire to walk
closer to the inner corner, have been considered in these studies. Almost all these approaches have captured

15-1
microscopic pedestrian movements through turning configurations qualitatively (i.e., path level). However,
it is uncertain whether they have captured microscopic behavioural dynamics quantitatively (i.e., with
regards to trajectories, speed and acceleration patterns). Although several studies have presented empirical
paths for turning pedestrians (for ex., [4] and [12]) no quantitative examination for speed and acceleration
patterns have been conducted. Thus, it is questionable that the movement mechanisms of pedestrians related
to walking through complex geometrical configurations have been adequately understood.
Reproducing realistic trajectories qualitatively and quantitatively, particularly associated with complex
geometries such as turning, in microscopic pedestrian models is beneficial in many ways. For example, to
accurately represent the bottleneck effect and for realistically displaying pedestrian movements in 3-D
visualizations particularly in virtual reality applications. Considering these benefits, this paper describes a
method using minimum-jerk theory to model trajectories of turning pedestrians. Data collected through
controlled experiments for different turning angles are utilized to validate the proposed approach.
The rest of this paper is structured as follows: The next section will discuss the theoretical background
of minimum-jerk theory. Then the methods, i.e., controlled experiments and modelling approach are briefly
described. This is followed by a qualitative and quantitative comparison of modelled and experimental
trajectories. Lastly, conclusions and recommendations for further studies are presented.

2. Optimal Trajectories and Minimum-Jerk Theory


Minimum-jerk concept has initially been utilized in neuroscience domain to study optimality
characteristics of skilled human arm movements. Flash and Hogan [13] verified that smoothness of a skilled
human planar arm movement, i.e., reaching, writing and drawing movements, can be evaluated as a function
of jerk (jerk is defined as the time derivative of acceleration). As they noted the objective function that
should be minimized to obtain the smoothest trajectory for moving hand from an initial position to a final
position in a given time T (T = tf – ti) is the time integration of the square of jerk that can be given as:
𝑡𝑓
2 2
1 𝑑3 𝑥 𝑑3 𝑦
J = ∫ (( 3 ) + ( 3 ) ) 𝑑𝑡 (1)
2 𝑑𝑡 𝑑𝑡
𝑡𝑖

Later, this theory was utilized to model more complex scenarios such as 2-dimentional robot arm
movements [14] and human reaching and catching movements in 3-dimentiolanal space [15]. Further, a
recent study revealed that turning manoeuvres of vehicles at intersections can be explained with minimum
jerk approach [16]. Applying the same theory, Pham et al. [17] demonstrated that whole body movement
share some common features with hand movements. That is, as they described, goal-oriented human
walking trajectories can also be described with minimum-jerk concept.
Through an experimental study, Dias et al. [18] described several behavioural characteristics that are
specific to walking through turning configurations. As they pointed out, pedestrians perform the turning
manoeuvre within a fixed region (described as “turning region”). Within this turning region there is a
deceleration phase followed by an acceleration phase and the minimum speed is occurred in the vicinity of
the middle of the corner. In this study we hypothesize that turning manoeuvres performed by individuals
within this turning region are skilled tasks and therefore, can be described with minimum-jerk approach
similar to reaching or drawing or goal-oriented movements.

3. Methods
3.1. Trajectory Data
Dias et al. [18] conducted a series of solo (one person at a time) walking experiments under different
conditions, i.e. for different turning angles (45°, 60°, 90°, 135°, 180°) and under 3 desired speed levels (i.e.
normal speed walking, fast speed walking and slow speed running). Out of those data collected through
these experiments, trajectories for normal speed walking through 90°, 135° and 180° turning configurations
were utilized in this study.

15-2
3.2. Modelling
As described in previous Section 2 the objective function to be minimized to obtain the minimum-jerk
trajectory is the time integration of squared jerk that is given in Equation 1. Note that the jerk is equivalent
to the rate of change of acceleration (or force) and therefore, a minimum jerk trajectory is a trajectory that
minimizes the rate of change of forces when changing initial walking direction gradually.
Solution for minimization of Equation 1 can be obtained as a system of fifth order polynomials as
given in Equations 2 and 3. Detailed derivations of this system of equation from equation 1, can be found
in [13].

𝑥(𝑡) = 𝑎0 + 𝑎1 𝑡 + 𝑎2 𝑡 2 + 𝑎3 𝑡 3 + 𝑎4 𝑡 4
(2)
+ 𝑎5 𝑡 5
𝑦(𝑡) = 𝑏0 + 𝑏1 𝑡 + 𝑏2 𝑡 2 + 𝑏3 𝑡 3 + 𝑏4 𝑡 4
(3)
+ 𝑏5 𝑡 5
Where;
𝑎𝑗 and 𝑏𝑗 (j = {0, ..., 5}) are constants

Equations 2 and 3 show that there are 12 unknowns and therefore 12 boundary conditions are required
to solve this system of equations. In this study, ti and tf (in Equation 1) are defined as the entry time to the
turning region and the exit time from the turning region respectively. For convenience, ti was set as zero.
Further, initial position (turn initiation point) was set as (0, 0) and the coordinates of the final positions
(turn completion point) for each turning angle case were set relative to the origin, based on the corridor
geometry based on the findings of [18]. Approaching speed at the turn initiation point and receding speed
at the turn completion point were set as 1.4 m/s as there were no statistically significant difference between
them as reported in [19]. Acceleration at initial point and final point was considered as 0 m/s 2 as it was
assumed, based on [18], that the deceleration and acceleration phases are occurred within the turning region.
Movement time from turn initiation point to turn completion point (T = tf – ti) is required to solve the above
system of equations. In Section 4 we consider that T is known and the average T for each angle case was
obtained from experimental trajectories. Later in Section 5 we show that T can be estimated as a function
of approaching speed and turning angle. For the above boundary conditions and average T obtained from
experimental trajectories, average minimum-jerk trajectories (paths, speed and acceleration profiles) were
obtained based on Equation 2 and 3 for each turning angle. These minimum-jerk trajectories are compared
with trajectories obtained from experiments as discussed in the next section.

4. Results
Modelled and experimental trajectories (walking paths, instantaneous walking directions, speed
profiles and acceleration profiles) were compared as depicted in Figure 1 and Figure 2. Absolute errors for
location, direction, speed and acceleration estimates were calculated according to Equation 4 and mean
absolute errors (MAE) for those estimates are tabulated in Table 1.

𝐴𝑏𝑠𝑜𝑙𝑢𝑡𝑒 𝑒𝑟𝑟𝑜𝑟 = |𝑆⃗𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑,𝑡 − 𝑆⃗𝐴𝑐𝑡𝑢𝑎𝑙,𝑡 | (4)


Where;
⃗S⃗Estimated,t = Estimated state vector at time t
⃗S⃗Actual,t = Actual (experimental average) state vector at time t

Table 1: Mean absolute errors of estimates.

Turning angle Path (m) Direction (°) Speed (m/s) Acceleration (m/s2)
90° 0.064 1.8 0.018 0.081
135° 0.075 3.0 0.025 0.065
180° 0.074 6.8 0.029 0.087

15-3
0.5 Experimental paths
Experimental average 180

Instantaneous walking direction (degrees)


Minimum-Jerk path
0
-0.5 0 0.5 1 1.5 2 2.5
X (m) 135
-0.5
Y (m)

-1 90

-1.5
45
Minimum-Jerk
-2
Experimental average
0
(a) 0.0 (d) 1.0 2.0 3.0 4.0
-2.5
0.5 Experimental paths Time (seconds)
180
Experimental average

Instantaneous walking direction (degrees)


Minimum-Jerk path
0
-0.5 0 0.5 1 1.5 2 2.5
135
X (m)
-0.5

90
Y (m)

-1

-1.5 45
Minimum-Jerk

Experimental average
-2
0
(b) 0.0 1.0 2.0 3.0 4.0
(e) Time (seconds)
-2.5
0.5 180
Experimental paths
Instantaneous walking direction (degrees)

Experimental average
Minimum-Jerk path
0 135
-0.5 0 0.5 1 1.5 2 2.5
X (m)
-0.5
90
Y (m)

-1

45

-1.5
Minimum-Jerk

Experimental average
0
-2 0.0 1.0 2.0 3.0 4.0
(c) (f) Time (seconds)

-2.5
Fig. 1: Comparison of walking paths and instantaneous walking directions for: (a) and (d) – 90°; (b) and (e) – 135°;
(c) and (f) – 180° turning movements (instantaneous walking direction is the angle between instantaneous
walking path and the x-axis measured clockwise direction).

15-4
1.6 0.8

1.2 0.4

Acceleration (m/s2)
0.8 0.0
Speed (m/s)

0.0 1.0 2.0 3.0 4.0


Time (seconds)
0.4 -0.4
Minimum-Jerk
(a) Experimental average (d)
0.0 -0.8 Minimum-Jerk
0.0 1.0 2.0 3.0 4.0 Experimental average
Time (seconds)
1.6 0.8

1.2 0.4

Acceleration (m/s2)
0.8 0.0
Speed (m/s)

0.0 1.0 2.0 3.0 4.0


Time (seconds)
0.4 -0.4
Minimum-Jerk
(b) Experimental average (e)
0.0 -0.8 Minimum-Jerk
0.0 1.0 2.0 3.0 4.0 Experimental average
Time (seconds)
1.6 0.8

1.2 0.4
Acceleration (m/s2)

0.8 0.0
Speed (m/s)

0.0 1.0 2.0 3.0 4.0


Time (seconds)
0.4 -0.4
Minimum-Jerk
(c) Experimental average (f)
0.0 -0.8 Minimum-Jerk
0.0 1.0 2.0 3.0 4.0 Experimental average
Time (seconds)
Fig. 2: Comparison of speed and acceleration profiles for: (a) and (d) – 90°; (b) and (e) – 135°;
(c) and (f) – 180° turning movements.

Table 1 explains that, in general, errors in estimates tend to increase with the increasing turning angle.
Path level comparison of modelled and experimental paths (Figure 1) suggests that the minimum jerk theory
can qualitatively predict the walking paths through turning configurations for different turning angles. Thus,
this could be useful in modelling desired direction of individuals through turning configurations.
Although speed and accelerations are slightly overestimated in modelled profiles (Figure 2) the trends,
i.e., acceleration and deceleration patterns, are well explained. In this study, speeds and acceleration values
for boundary conditions were approximated. For example, initial and final accelerations at turn initiation
point and turn completion points were assumed to be zero. However, experimental average profiles show
that initial and final accelerations are not exactly zero. Provision of such complete boundary conditions or
using higher order models may generate more accurate trajectories. Further, individual variations of
boundary conditions and movement times were also not considered in this study. Regardless of these
approximations, trajectories could be generated with a reasonable accuracy using the minimum jerk
approach.

15-5
5. Variations in Turning Trajectories
Properties of trajectories are dependent on boundary conditions, particularly on entry and exit speed
as well as on the movement time through the turning region. For the analyses so far discussed in Section 4,
it was considered that the movement time (T = tf – ti) through the turning region is known. However, in
applications, such as in microscopic simulation models, T is generally unknown. This section will briefly
discuss such issues combined with trajectory variations through turning configurations.

5.1. Properties of Movement Time (T)


Figure 3 plots the variation of T with respect to the approaching speed and the turning angle. As can
be explained from this figure T is basically dependent on the approaching speed to the turning region and
turning angle. Therefore, T could be simply modelled as functions of those variables. This provides a
reasonable estimate for T as further discussed in Section 5.2. It should be noted that, as described in [19],
approaching speed to the turning region does not vary with the turning angle. Or in other words,
approaching speeds for different turning angles do not display statistically significant differences among
them.

5.0
Movement time through turning region (s)

4.5 180°
y = -1.8939x + 6.7386
R² = 0.4537

4.0 135°

90°
3.5 y = -2.1829x + 6.6074
R² = 0.36

3.0

2.5
y = -2.0045x + 6.0104
R² = 0.8867
2.0
1 1.2 1.4 1.6 1.8
Approaching speed (m/s)

Fig. 3: Variation of movement time through the turning region with approaching speed and turning angle.

5.2. Trajectory Variation


Variation of trajectories with respect to approaching speed to the turning region are explored in this
section. In this analysis, distributions of approaching speeds to different turning angles were considered
and the trajectory variation for slow (5th percentile of approaching speeds) and fast (95th percentile of
approaching speeds) walkers were investigated. Movement time for each approaching speed and turning
angle was determined based on the linear relationships shown in Figure 3. Exit speed from the turning
region was considered to be similar to the approaching speed. Estimated 5th percentile and 95th percentile
approaching speeds for all angle cases were [1.2, 1.3] m/s and [1.55, 1.65] m/s respectively. Similar to
settings in Section 4, accelerations at the entry and exit boundaries of turning region were set as 0 m/s 2. For
such settings and approximations, trajectories were generated for each turning angle case and compared as
depicted in Figure 4 and 5. Figure 4 describes that, although deviations are smaller, variation in walking
paths tend to increase with increasing turning angle. Regarding speed and acceleration profiles, increase in
variations in trajectories with increasing turning angle can be clearly observable in Figure 5.

15-6
0.5 High approaching speed (95th percentile)
Low approaching speed (5th percentile)
Average approaching speed
0
-0.5 0 0.5 1 1.5 2 2.5
X (m)

-0.5

-1

Y (m)
-1.5
180°

-2
90°
135°
-2.5

Fig. 4: Variation of walking paths through different turning angles based on different approaching speeds.

0.8

1.6 90° 90°


0.4
Acceleration (m/s2)
Speed (m/s)

1.2
0.0
0.0 1.0 2.0 3.0 4.0
0.8 Low approaching speed
-0.4 Low approaching speed
Average approaching speed
Average approaching speed
High approaching speed
0.4 High approaching speed
-0.8
0.0 1.0 2.0 3.0 4.0
Time (seconds)
Time (seconds)
0.8

1.6
135° 135°
0.4
Acceleration (m/s2)
Speed (m/s)

1.2
0.0
0.0 1.0 2.0 3.0 4.0
0.8
Low approaching speed -0.4 Low approaching speed
Average approaching speed Average approaching speed
High approaching speed High approaching speed
0.4
-0.8
0.0 1.0 2.0 3.0 4.0
Time (seconds)
Time (seconds)
0.8

1.6
180° 180°
0.4
Acceleration (m/s2)
Speed (m/s)

1.2
0.0
0.0 1.0 2.0 3.0 4.0

0.8
Low approaching speed -0.4 Low approaching speed
Average approaching speed Average approaching speed
High approaching speed High approaching speed
0.4 -0.8
0.0 1.0 2.0 3.0 4.0
Time (seconds) Time (seconds)
Fig. 5: Variation of speed and acceleration patterns through different turning angles based on different
approaching speeds.

15-7
It can be noted that, for a given turning case, although walking paths do not display larger deviations
(for ex., consider 90° case), speed and acceleration patterns can be clearly distinguished for different
approaching speeds. This observation highlights that individuals might adjust their speed and acceleration
patterns as well to negotiate a corner smoothly. Therefore, although the walking path is the same, speed
and acceleration behaviours could be considerably different for different approaching and exit conditions.
Further, it can be observable in Figure 5 that deceleration and acceleration values are larger for higher
approaching speeds and this effect is magnified when the turning is increased. Pedestrians approaching a
corner with higher speeds have larger inertia. In order to overcome the effect of this inertia and negotiate
the turning movement smoothly, a pedestrian requires a larger deceleration force compared to a pedestrian
approaching at a lower speed. Then, for an individual leaving at higher speed (approaching and leaving
speeds were considered same in this section), a larger acceleration is required to regain the speed to the
desired level. Such behaviours have adequately captured with minimum-jerk approach as described in this
study.
Findings of this study reveal important information related to mechanisms and optimality
characteristics of human movements through complex environments. Minimizing jerk is equivalent to
minimizing the rate of change of acceleration (or force). Thus, it can be stated, based on the findings of this
study, that human tend to minimize the rate of change of acceleration when navigating through turning
configurations. In the future it may be possible to consider such evidences in microscopic pedestrian
simulation models to realistically model pedestrian movement characteristics.

6. Conclusions
Understanding microscopic behavioural dynamics and mechanisms associated with walking through
complex geometrical configurations could be useful particularly for enhancing or developing microscopic
simulation models. In this paper, minimum jerk theory was applied to model trajectories for walking
through turning configurations at normal walking speed. Through a qualitative and quantitative comparison
of predicted and experimental trajectories, it was clarified that the minimum jerk theory can adequately
predict microscopic walking characteristics, for ex., walking paths and acceleration-deceleration patterns,
through turning configurations under different conditions (for different angles and different approaching
speeds). Variation of trajectories were also tested based on different conditions (different approaching
speeds and movement times) and verified that the model is responsive and sensitive to such different
movement conditions.
Further, this study confirmed that people tend to optimize the smoothness of their movements through
complex geometrical settings by minimizing the jerk. Such concept, which describe realistic movement
mechanisms of people through turning configurations, could be used to model and simulate complex
pedestrian movements more realistically.
As further investigations, a comprehensive sensitivity analysis should be performed to better
understand the uncertainty of predictions via minimum jerk theory. Further, more complex scenarios related
to turning configurations could also be explored under a wide variety of boundary conditions.

References

[1] S. Hoogendoorn, M. Hauser, and N. Rodrigues, "Applying microscopic pedestrian flow simulation to
railway station design evaluation in Lisbon, Portugal," Transportation Research Record: Journal of the
Transportation Research Board, vol. 1878, pp. 83–94, 2004.
[2] C. Rogsch, W. Klingsch, A. Seyfried and H. Weigel, “Prediction Accuracy of Evacuation Times for
High-Rise Buildings and Simple Geometries by Using Different Software-Tools,” in Traffic and
Granular Flow ’07, C. Appert-Rolland, F. Chevoir, P. Gondret, Ed. Springer Berlin Heidelberg, 2009,
pp. 395-400.
[3] D. C. Duives, W. Daamen, and S. P. Hoogendoorn, "State-of-the-art crowd motion simulation
models," Transportation Research Part C: Emerging Technologies, vol. 37, pp. 193–209, 2013.

15-8
[4] B. Steffen and A. Seyfried, “Modeling of pedestrian movement around 90 and 180 degree bends,” in
Proceedings of the Workshop on Fire Protection and Life Safety in Buildings and Transportation
Systems, J.A. Capote, Alvear, D., Ed. Universidad de Cantabria, GIDAI, Santander, 2009.
[5] J.-B. Zeng, B. Leng, Z. Xiong, and Z. Qin, “Pedestrian dynamics in a two-dimensional complex scenario
using a local view field model, “International Journal of Modern Physics C, vol. 22, no. 08, pp. 775–
803, 2011.
[6] C. Dias, O. Ejtemai, M. Sarvi, and M. Burd, "Exploring pedestrian walking through Angled
corridors," Transportation Research Procedia, vol. 2, pp. 19–25, 2014.
[7] M. Chraibi, M. Freialdenhoven, A. Schadschneider, and A. Seyfried, “Modeling the Desired Direction
in a Force-Based Model for Pedestrian Dynamics,” in Traffic and Granular Flow ’11, V. V. Kozlov, A.
P. Buslaev, A. S. Bugaevet, Ed. Springer Berlin Heidelberg, 2013, pp. 263-275.
[8] X.-X. Jian, S. C. Wong, P. Zhang, K. Choi, H. Li, and X. Zhang, "Perceived cost potential field cellular
automata model with an aggregated force field for pedestrian dynamics," Transportation Research Part
C: Emerging Technologies, vol. 42, pp. 200–210, 2014.
[9] T. Kretz, “The use of dynamic distance potential fields for pedestrian flow around corners,” in First
International Conference on Evacuation Modeling and Management (ICEM 09), TU Delft, the
Netherlands, 2009.
[10] R.-Y. Guo and T.-Q. Tang, "A simulation model for pedestrian flow through walkways with
corners," Simulation Modelling Practice and Theory, vol. 21, no. 1, pp. 103–113, 2012.
[11] S. Li, X. Li, Y. Qu, and B. Jia, "Block-based floor field model for pedestrian’s walking through
corner," Physica A: Statistical Mechanics and its Applications, vol. 432, pp. 337–353, 2015.
[12] F. Johansson, D. Duives, W. Daamen, and S. Hoogendoorn, "The many roles of the relaxation time
parameter in force based models of pedestrian dynamics," Transportation Research Procedia, vol. 2,
pp. 300–308, 2014.
[13] T. Flash and N. Hogan, "The coordination of arm movements: an experimentally confirmed
mathematical model," The journal of Neuroscience, vol. 5, no. 7, pp. 1688–1703, 1985.
[14] Y. Wada and M. Kawato, "A via-point time optimization algorithm for complex sequential trajectory
formation," Neural Networks, vol. 17, no. 3, pp. 353–364, 2004.
[15] M. Bratt, C. Smith, and H. Christensen, “Minimum jerk based prediction of user actions for a ball
catching task,” in Intelligent Robots and Systems, 2007. IROS 2007. IEEE/RSJ International Conference
on, 2007, pp. 2710-2716.
[16] C. Dias, M. Iryo-Asano, and T. Oguchi, “Concurrent prediction of location, velocity and acceleration
profiles for left turning vehicles at signalized intersections,” in Proceedings of 53th Annual Meeting of
Infrastructure Planning (JSCE), 2016, Hokkaido, Japan.
[17] Q.-C. Pham, H. Hicheur, G. Arechavaleta, J.-P. Laumond, and A. Berthoz, "The formation of
trajectories during goal-oriented locomotion in humans. II. A maximum smoothness model," European
Journal of Neuroscience, vol. 26, no. 8, pp. 2391–2403, 2007.
[18] C. Dias, O. Ejtemai, M. Sarvi, and N. Shiwakoti, "Pedestrian walking characteristics through Angled
corridors: An Experimental Study," Transportation Research Record: Journal of the Transportation
Research Board, vol. 2421, pp. 41–50, 2014.
[19] C. Dias, M. Sarvi, N. Shiwakoti, and O. Ejtemai, “Experimental study on pedestrian walking
characteristics through angled corridors,” in Proceedings of 36th Australasian Transport Research
Forum (ATRF), 2013, Brisbane, Australia.

15-9

View publication stats

You might also like