This paper presents two contributions in the development of stair detection algorithm for climbing robot. First, a new tri-star wheeled mechanism was developed to smoothly overcome the stairs. Second, the sensor-based algorithm was proposed for detection of stairs and switching states autonomously. Two motor driven ultrasonic sensors feed the posture information of the robot back to the controller to recognize the climbing environment. The validity of the proposed algorithm was demonstrated through experiments in realizing climbing environment. The experiments prove that the proposed algorithm can do rolling, ascending and descending on the staircase.

1. Stairs Detection Algorithm for Tri-Star Wheeled Robot and Experimental Validation
Stairs Detection Algorithm for Tri-Star Wheeled Robot
and Experimental Validation
Aye Mya Mya Thu
Department of Mechatronic Engineering, Yangon Technological University, East Gyogone, Insein, Yangon, Myanmar.
E-Mail: ayemyathu579@gmail.com, Tel: +95 9970353549.
This paper presents two contributions in the development of stair detection algorithm for climbing
robot. First, a new tri-star wheeled mechanism was developed to smoothly overcome the stairs.
Second, the sensor-based algorithm was proposed for detection of stairs and switching states
autonomously. Two motor driven ultrasonic sensors feed the posture information of the robot
back to the controller to recognize the climbing environment. The validity of the proposed
algorithm was demonstrated through experiments in realizing climbing environment. The
experiments prove that the proposed algorithm can do rolling, ascending and descending on the
staircase.
Keywords: Climbing robot, Detection algorithm, Mechanism design, Range sensing, Sensor-based control.
INTRODUCTION
To cope with the climbing environment in the real world,
various research efforts have been directed toward the
development of a robot capable of autonomous stair
climbing. Staircase is a common obstacle that a mobile
search and rescue robot may encounter during every
mission. Climbing up and down stairs for a mobile robot is
not an easy task. Zhang et al. (2011) developed a laser
based algorithm where a vertical laser and sonar sensor
are used to identify stairs and a fuzzy control system is
developed to ensure that the robot does not stray from the
centerline of the stairs during approaching the staircases.
The autonomous stair climbing algorithm was implemented
on tracked robot by Vu et al. (2008), Pinhas et al. (2007),
Shi et al. (2017) and Morozovsky et al. (2015) which is
based on multi-sensor fusion and camera detection. Cong
et al. (2007) implemented a stair detection method
implemented on UGV with the innovation of a control
algorithm using Kinect v2 depth sensor to recognize and
measure stairs for stair climbing tasks of tracked robot.
Michael et al. (2014) proved that the wheelchair is the
solution for the demographic change in society as it
promises the maximum of mobility for people who suffer
physical impairment due to a stair climbing function and a
car integration. Kinematics based approach has been
proposed for rescue robot in the performance of Kalantari
et al. (2009), in order to improve their motion control and
pose estimation. Lai et al. (2010) conducted the
development of an image based cross-floor navigation
method combined with wireless modules. Lai et al. (2009)
investigated methods to detect and locate curbs and
stairways using stereo vision. The previous researches are
very motivated for the climbing environment but the
recognitions of climbing environment are not so simple for
those who unfamiliar with image processing. To simply
cope with the above complex techniques, safe and
autonomous stair climbing algorithm is proposed and
implemented for a new tri-star wheeled robot in this paper.
Aiming at the functional requirement of climbing the stairs,
Aye et al. (2016) has been reported on dynamics and
kinematics analysis during a tri-star wheeled mobile robot’s
climbing of stairs. Simulation results are also provided that
the calculations are valid. This study aims to develop
simple sensor-based stair detection algorithm combined
with the reliable control strategy to switch rolling,
ascending and descending states. To climb the stair
autonomously, two motor driven ultrasonic sensors feed
the posture information of the robot back to the controller.
The proposed algorithm is based on the sensors’
information during approaching and landing. As a smooth
Research Article
Vol. 4(2), pp. 041-048, August, 2019. © www.premierpublishers.org. ISSN: 1550-7316
World Journal of Mechanical Engineering

2. Stairs Detection Algorithm for Tri-Star Wheeled Robot and Experimental Validation
Aye MMT 042
Figure 1: Sensors and actuators demonstration of the
proposed tri-star wheeled robot. (Source: Author)
outcome, it completes the switching state to the robot. To
ascend and descend the stairs with the simplest way is the
motivation of this paper but not to reduce the number of
sensors and actuators. The main contribution of
mechanism design is the use of only one actuator to drive
both of wheel rotation and tri-star frame rotation. The
limitation is that the proposed system is appropriate for the
proposed stair. This means that if other stairs design is
approached, the successive probability of proposed
algorithm may be less effective although not absolutely
useless.
APPROACH TO PROPOSED SYSTEM
This section presents a brief introduction to the proposed
system. The mechanism design used in the experiment is
very simple and the focus is to use only one actuator
rotating forward for rolling and backward for climbing. By
redirecting the function of reversing the actuator to
climbing, it loses the capability of rolling the vehicle
backward with a single actuator. Actually, it is the strong
point of the proposed contribution although it seems
drawback because falling backward condition can be
prevented during the climbing motion.
If there is any request for desired application areas, the
other movements like backward and turning can be
achieved by the additional derivative set of mechanism and
the development of control algorithm. Structure
components of the proposed design is demonstrated in
Figure 1. Due to the complexity of autonomous stair
climbing, the procedure is then decomposed into a
sequence of three individual tasks with the additional
stages as illustrated in Figure 2. The proposed design
incorporates a number of sensors and actuators, which
leads the robot to simply switch among rolling, ascending
and descending. The proposed system is not only to
operate on the flat surface like the conventional wheels but
also to drive on the inclined surface in order to overcome
the stairs. Therefore, two ultrasonic sensors are needed to
be controlled by the two individual servomotors because
they must be always in the predefined positions providing
for the proposed algorithm. Besides, two motor driven
ultrasonic sensors will give the stable output just with the
use of interrupt pin for two sensors and the power for two
servomotors are individually given by the motor driver. The
other considerations will give the unstable outcomes as
shown in Figure 3. Due to the requirement of predicting the
deviation of ultrasonic sensors, one gyro sensor is used to
detect the inclination angle of the robot’s body. In this way,
two ultrasonic sensors will be reached back to the original
position every time when they are conflicted. Unlike the
conventional robots, the tri-star wheeled robot can perform
two rotations: wheel rotation leading the robot to rolling and
tri-star frame rotation leading the robot to ascending and
descending. Thus, two encoders are used to individually
measure two rotations. Here, it is needed to note that both
of rolling angle and tri-star angle are driven by the same
actuator. So, it is reliable that rolling angle is not rotating
altogether with the rotation of tri-star frame angle although
the vice versa is operated.
Figure 2: Flow chart of the proposed algorithm.
(Source: Author)

3. Stairs Detection Algorithm for Tri-Star Wheeled Robot and Experimental Validation
World J. Mech. Engin. 043
An additional check of tri-star frame rotation has provided
to confirm that the robot can safely land on the stair surface
and can assist to the detection of two sensors before
changing to the next state. Although the two sensors are
always updating the information, the changes are valid just
after landing the tri-star frame. The relevant velocity is also
applied to the robot for each state heading to have the
smarter and safer sequence for the system.
AUTONOMOUS STAIR DETECTION ALGORITHM
This section describes the design and implementation of
the stair detection algorithm for the proposed robot. The
objectives and approaches for the individual stages will be
introduced respectively with the emphasis on range
detection of the two motor driven ultrasonic sensors. The
effectiveness of sensor ranges on different stair sizes
leading to optimize the appropriate setting of sensors’
positions and height of staircases. The contribution of
algorithm is detailed in Algorithm 1, where the notations are
adopted as follows: d is the distance between the
horizontal ultrasonic sensor and the step, α is the threshold
value for d, x is the inclined distance between the inclined
ultrasonic sensor and the step, and γ is the threshold value
for x. Then, θ is the rotation angle of tri-star frame. These
variables are input to the controller and as the output the
robot will do rolling, ascending and descending. Initially,
the robot will start with rolling meaning that tri-star frame
switch denoted as sT and speed down switch denoted as
sSD are off. The condition of inclined sensor will be always
in the state of x <=γ while facing with the upstairs.
Figure 3: Experimental results of two motor driven
ultrasonic sensors: (a) with interrupt pin and with power
from motor driver, (b) with interrupt pin and without power
from motor driver, (c) without interrupt pin and with power
from motor driver, (d) without interrupt pin and without
power from motor driver. (Source: Author)
For the horizontal sensor, if α+1 <= d <= α +2 is detected,
the robot has to change a low forward velocity before
accelerating to the desired ascending velocity during
closing to the stairs to prevent the mechanism from
bumping about the first step. So, sSD will be on. If the value
of d is absolutely equal to the threshold value of α, switch
sT will on and the rolling state will switch to ascending state.
Figure 4: Demonstration of the proposed algorithm.
(Source: Author)
As soon as the tri-star frame rotation is started, it will be
always recorded and when reach to the multiples of 120
degrees, the information of two sensors will be checked
again to give the switch decision to the robot. In switching
to descending state, switch sT do the same function with an
ascending state because the tri-star frame rotation makes
the robot both ascending and descending. The result is
completely depended on the distance changes of two
sensors. In this state, both of x and d are required to be
greater than the threshold values to get descending state.
If not, rolling state will be occupied. Furthermore, activation
of wheel rotation results in rolling state and

4. Stairs Detection Algorithm for Tri-Star Wheeled Robot and Experimental Validation
Aye MMT 044
ascending/descending state outcomes from activation of
tri-star frame rotation. Demonstration of the proposed
algorithm is shown in Figure 4.
SENSORS AND ACTUATORS
The challenge of proposed algorithm is to identify and
conquer the individual steps with clearly defined switching
states. To overcome the complete computation of
staircase’s parameters, only the information from two
ultrasonic sensors are used to predict the existence of stair
on ahead of. The demonstration of two ultrasonic sensors
is illustrated in Figure 5. The strategy to control the sensors
and actuators is very important to give the complete
information to the controller and to provide for the switching
decision of the proposed algorithm. Ascending and
descending the stairs can be divided into three main tasks.
(1) Approaching, (2) Climbing and (3) Landing. To be
always kept the two sensors in leading ahead is tuned by
two servomotors during approaching and after landing. Tri-
star frame angle rotation is undertaking after approaching
to complete climbing.
Figure 5: Demonstration of sensors and actuators.
(Source: Author)
1) Prediction about Upstairs: The first ultrasonic sensor
is implemented in the horizontal direction to be parallel
with the ground level. This one is intended to detect the
condition that there is the stair or not in front of the
robot leading to know before the upstairs. If not the
mechanism can be damaged by bumping with the
staircase. To prevent from this condition, the horizontal
ultrasonic sensor is needed to let the robot know
before the mechanism touches with the stair.
2) Prediction about Downstairs: The second ultrasonic
sensor is mounted in the inclined position to be parallel
with the tri-star frame instead of horizontally. This is
because it is intended for the prediction of downstairs.
Unlike with the upstairs, downstairs are located in the
inclined position and they cannot be detected with
horizontal sensor. This can lead to the falling down of
the robot. Implementing the second sensor in incline
position can solve these problems. Therefore, the
robot will know before the downstairs.
3) Detection of the Inclination of Robot’s Body: For
the heading control of two ultrasonic sensors, a simple
control algorithm is implemented. Utilizing the control
algorithm and with gyro feedback, experimental results
show that the sensors are able to maintain the desired
heading while the robot is on inclined surface, and give

5. Stairs Detection Algorithm for Tri-Star Wheeled Robot and Experimental Validation
World J. Mech. Engin. 045
the correct information to the controller if there are no
significant effects on the robot heading. Significant
heading errors occur when the climbing environment is
rough or disturbed by other obstacles except the steps.
These errors are normal for the sensors as their
feedback is based on the sound signal. Experimental
results in Figure 6 and Figure 7 show that the robot is
able to overcome and continue the stair climbing with
desired switching state using the proposed algorithm.
4) Specifications of Sensors and Actuators: The
configurations related to the devices used in the
proposed system are described in the following table.
Sensors and
actuators Descriptions
Ultrasonic
sensor
Supply voltage : 5 V
Current : 15 mA
Frequency :40 kHz
Maximal range : 400 cm
Minimal range : 3 cm
Trigger pulse width: 10 μs
Gyro sensor Power supply : 4.3 to 9 V
Gyroscope range: +/-250, 500, 1000, 2000 °/s
Acceleration range: +/-2, 4, 8,16 g
Weight : 2.1 g (0.07oz)
Servo motor Operating voltage : 4.8-6.0 V
Operating speed (4.8V): 0.23sec/60 degrees
Operating Speed (6.0V): 0.19sec/60 degrees
Stall Torque (4.8V) : 44 oz/in (3.2kg.cm)
Stall Torque (6.0V) : 56.8 oz/in (4.1kg.cm)
Weight : 1.3 oz (37.2g)
Actuator Nominal voltage: 12 V
Gear ratio : 100 : 1
Stall current : 6 A
Output torque : 1.55 Nm
Rotary encoder : 64 CPR
Motor power : 12 V
No load speed :100 rpm
Weight : 230 g
DETECTION AND APPROACH
The main motivation of using ultrasonic sensors is to
complete the stairs detection process without calculating
the complete parameters of staircases (i.e, height, width,
length, inclination). Two sensors will be simultaneously
detected the stair and proposed algorithm will assist for the
robot to decide the stair on ahead of is upstairs or
downstairs. Then, the tri-star frame rotation will also be
always checked. Every time when it reaches to the
multiples of 120 (0, 120, 240, 360) degree, the information
of two ultrasonic sensors will be checked again for the
switching states.
1) Detection of Ascending State: As the actuator is
started to activate, the two servomotors are rotated to
control the orientation of two sensors, as demonstrated
in Figure 6(b) and Figure 6(c), according to the
changes of inclination angle of robot’s body as shown
in Figure 6(a). If the state of tri-star frame angle is
satisfied with the multiples of 120 degrees, the first
consideration is to check the information from
horizontal sensor. If d > γ, it means the robot is not
close up to the step. But only this information is not
possible for the robot to have the accurate decision.
Therefore, the second consideration is added to
ensure that the inclined sensor is less than or equal to
the threshold value (x <= γ). If so, the robot will be on
flat surface with the normal velocity defined for rolling
state. Here, the tri-star frame angle, θ will be zero and
only the rolling angle, β will be activated since it is the
rolling state. In this situation, the gyro feedback is not
considered as the two sensors are in initial states.
Since the robot is in rolling state, the information from
two sensors are always changes and when the range
of d is within the deviation of increment or decrement
of 2 compared to the threshold value, the previous
velocity will be decreased in order to prevent the
mechanism from bumping into the steps. After 4 sec,
the data in Figure 6(d) and Figure 6(e) show that the
actual distance of ultrasonic sensor 1, d becomes
equal to the threshold value of α and the actual
distance of ultrasonic sensor 2, x becomes greater
than γ then the velocity will be increased enough to
climb the step because ascending state require the
higher torque than rolling state and as a result the
rolling mode is switched to the ascending mode as
shown in Figure 6(g) and the descending mode is off
during the ascending mode is on as can be seen in
Figure 6(h). Once the ascending mode is on, the tri-
star frame is rotated starting from zero as shown in
Figure 6(f), and at the same time the rolling angle, β is
simultaneously rotated altogether with the rotation of θ
because of the mechanism design as can be seen in
Figure 6(i). After landing on the first step, the gyro
feedback is considered to recover the deviation of two
sensors. Similar actions will be activated when the next
step is approached. To satisfy the detection of upstairs
and to ascend each step, just 120 degrees’ rotation of
tri-star frame is needed. Therefore, as confirmed in
Figure 6(f), one revolution and 120 degrees of tri-star
frame angle is required to overcome four steps. So,
counter-clockwise rotation of the actuator gives the tri-
star motion.
2) Detection of Descending State: For the downstairs,
the information of inclined ultrasonic sensor plays in a
vital role to let the robot know before the steps and to
prevent from falling down. Initially, the robot is at the
top of the downstairs and the distance of horizontal
ultrasonic sensor is always greater than the threshold
value. As the initial state of descending is rolling mode,
the condition of d > γ, x <= γ, θ = 0 have been satisfied
and β will be activated with the rolling velocity. The
approach to downstairs is straightforward with the

6. Stairs Detection Algorithm for Tri-Star Wheeled Robot and Experimental Validation
Aye MMT 046
Figure 6: Experimental results of switching state between Figure 7: Experimental results of switching state
rolling mode and ascending mode based on such data between rolling mode and descending mode based on
of λ, μ, σ, d, x, θ and β. (Source: Author) such data of λ, μ, σ, d, x, θ and β. (Source: Author)
desired position of two sensors and the gyro feedback
is not affected. The more the robot is close up to the
down steps, the greater the range of inclined sensor. If
x > γ is satisfied, the rolling velocity is needed to be
decreased to safely overcome the steps because
descending state does not require the higher torque as
in ascending state. Then, the descending state will be
switched smoothly. This means that wheel rotation is
changed to tri-star frame rotation. Thus, as in the
ascending state, tri-star frame angle is initiated to
rotate starting from zero simultaneously with the
activation of rolling angle. Throughout the changes of
desired position, the two sensors will orientate
themselves to the desired heading, based on the
information obtained from the gyro sensor. To
overcome a down step and to complete landing, the
triangle position of tri-star frame which means 240
degrees is needed to rotate. Since the proposed stair
has four steps, the tri-star frame is needed to rotate
two revolutions and 240 degrees for the whole
scenario. This was confirmed in Figure 7(f) with the
instantaneous rotation of rolling angle as can be seen
in Figure 7(i).
3) Declaration of Nomenclatures: The nomenclatures
used in the paper can be denoted as follows:
λ = inclination angle of the robot’s body,
μ = rotation angle of servo motor 1 with respect to λ,
σ = rotation angle of servo motor 2 with respect to λ,
d = actual distance of ultrasonic sensor 1,
x = actual distance of ultrasonic sensor 2,
θ = tri-star frame angle,
β = rolling angle.
4) Range Limitations for Different Stairs Size: The
robot first needed to find the scope of untouchable
area, see in Figure 8, where the area must not meet
the staircase surface and its scope can be adjusted
according to the height of staircase. The adjustable
ones are the postures of two ultrasonic sensors and
two servomotors which are denoted as h1 and h2.
Figure 8: Setting of untouchable area for different stairs
size. (Source: Author)
Normally, the robot can overcome the height of step
that is equal to the radius of the tri-star frame. The two
sensors can provide the valid information for the
proposed stair design. So, the other conditions should
also be considered about the stairs different from the
proposed one. For the lower step, the robot will
overcome with the rolling state instead of ascending
state because the height of step is too low for two
sensors to capture the distances. If the step is a little
lower than the proposed one that means the horizontal
sensor captures the fluctuated distance, the stair
climbing probability of robot will depend on the landing

7. Stairs Detection Algorithm for Tri-Star Wheeled Robot and Experimental Validation
World J. Mech. Engin. 047
position of tri-star frame. If the fluctuated signal is got
while the tri-star frame is out of the multiples of 120
degrees, the input signal to the controller is not
effective to the switching state. Otherwise, the wrong
information can conflict to the state decision. For the
higher step, the robot is sure not to overcome the step.
If the tri-star frame is fortunately flip over the step, the
inclined sensor will be bumping to the step surface. So,
the most satisfaction step for the proposed robot is the
height between the upper level of horizontal ultrasonic
sensor and the center of tri-star frame. Figure 9 shows
the demonstration of range limitations for two
controlled ultrasonic sensors.
Figure 9: Range limitations of two controlled ultrasonic
sensors on different stairs size. (Source: Author)
5) Optimization of stair climbing ability: Depending on
the radius of tri-star frame and the scope of
untouchable area, the climbing probability can be
optimized for different heights. The conditions of q = 0
and q = b are physically impractical. The scope of
untouchable area, q must be between these two
parameters. For 0 < q < b, the best position can be
selected for different a and different q. Everyone can
define all of parameters with different ways form the
proposed deign. But, after testing for a lot of times, the
most probability is adopted in Figure 10.
Figure 10: Optimization of robot’s climbing ability based on
H, q and r0. (Source: Author)
CONCLUSION
In this paper, a complete stair climbing up and down
scheme has been proposed where the tasks involve
detection of stairs, preparation stage, actual ascending,
actual descending and subsequent landing. It is especially
verified that the use of two directional control ultrasonic
sensors can provide the better stair detection and the safe
descending state instead of the use of one sensor.
Moreover, sensing the stairs with two sensors can easily
recognize the climbing environment. The proposed
algorithm is clearly detailed in pseudo code and also with
the demonstration layout. Experimental results have been
presented to validate the subsequent performance of the
proposed algorithm. The algorithm developed in this paper
was tested in a number of scenarios and has been proven
to perform successfully on the proposed staircase. The
proposed system provides the operation of one set of tri-
star on ascending and descending staircase. The
additional sets are needed to constructed with the
development of switching algorithm for backward and
turning movements.
REFERENCES
Aye, M.M.T., Soe, T.Z. and Okada, T. (2016). Dynamic
analysis for both rolling and climbing of tri-star wheeled
robot. International Organization of Scientific Research
Journal, 13(5), 52-62.
Cong, Y., Li, X., Liu, J. and Tang, Y. (2007). A stairway
detection algorithm based on vision for UGV stair
climbing. Proceedings of the IEEE International

8. Stairs Detection Algorithm for Tri-Star Wheeled Robot and Experimental Validation
Aye MMT 048
Symposium on Safety, Security and Rescue Robotics.
1806-1811. DOI: 10.1109/ROSE.2007.4373976.
Kalantari, A., Mihankhah, E. and Moosavian, S.A.A.
(2009). Safe autonomous stair climbing for a tracked
mobile robot using a kinematics based controller.
IEEE/ASME International Conference on Advanced
Intelligent Mechatronics, Singapore. 1891-1896.
DOI: 10.1109/AIM.2009.5229765.
Lai, W.M. and Lin, C.Y. (2009). Autonomous staircase
detection and stair climbing for a tracked mobile robot
using fuzzy controller. Proceedings of the IEEE
International Conference on Robotics and Biomimetic,
Thailand. 1980-1985. DOI: 10.1109/ROBIO.2009.
4913304.
Lai, W.M. and Lin, C.Y. (2010). Autonomous cross-floor
navigation of a stair climbing mobile robot using
wireless and vision sensor. IEEE/ASME International
Conference on Advanced Intelligent Mechatronics,
Singapore. 1971-1977. DOI: 10.1109/ISR.2013.
6695649.
Michael, H., Petra, F., Eichinger, A. and Wolf, B. (2014).
Stair sensing system based on optical 3D data for an
autonomous stair-climbing wheelchair. IEEE Fourth
International Conference on Consumer Electronics
Berlin. 400-403. DOI: 10.1109/ICCE-Berlin.2014.
7034274.
Morozovsky, N. and Bewley, T. (2015). Stair climbing via
successive perching. IEEE/ASME Transactions on
Mechatronics. 20(6), 1-10. DOI: 10.1109/TMECH.2015.
2426722.
Pinhas, B.T., Shingo, I. and Andrew, A.G. (2007).
Autonomous stair climbing with reconfigurable tracked
mobile robot. IEEE International Workshop on Robotic
and Sensors Environments, Canada. DOI:
10.1109/ICNSC.2008.4525517.
Shi, J.G., Zhu, W. and Wang, J. (2017). Approach to
autonomous stair climbing for tracked robot.
Proceedings of the IEEE International Symposium on
Safety, Security and Rescue Robotics. 182-186.
DOI: 10.1109/ICUS.2017.8278337.
Vu, Q.H., Kim, B.S. and Song, J.B. (2008). Autonomous
stair climbing algorithm for a small four-tracked robot.
International Conference on Control, Automation and
Systems, Korea. 2356-2360. DOI: 10.1109/ICCAS.
2008.4694199.
Zhang, Q., Ge, S.S. and Tao, P.Y. (2011). Autonomous
stair climbing for mobile tracked robot. Proceedings of
the IEEE International Symposium on Safety, Security
and Rescue Robotics. 92-98. DOI: 10.1109/SSRR.
2011.6106757.
Accepted 5 August 2019
Citation: Aye MMT. (2019). Stairs Detection Algorithm for
Tri-Star Wheeled Robot and Experimental Validation.
World Journal of Mechanical Engineering 4(2): 041-048.
Copyright: © 2019: Aye MMT. This is an open-access
article distributed under the terms of the Creative
Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium,
provided the original author and source are cited.