Professional Documents
Culture Documents
Research paper
a r t i c l e i n f o a b s t r a c t
Article history: Wheeled rovers have a limited ability to traverse soft terrains and climb loose slopes. This
Received 6 February 2018 study proposes a wheel-step rover with a modified active rocker-bogie suspension. The
Revised 4 May 2018
rocker and differential shaft fixed to the rocker are divided into three parts and recon-
Accepted 4 May 2018
nected via a novel angle-adjusting mechanism. The mechanism, which is driven by a single
motor, widens and narrows the angle between the two rocker parts periodically to gener-
Keywords: ate a wheel-step motion. Moreover, it adjusts the angle between a rocker part and differ-
Angle-adjusting mechanism ential shaft part to keep the bottom side of the rover body parallel to the ground. When
Planetary gear train fully stretched, the suspension lowers the bottom side horizontally to the lander platform.
Active suspension The mechanism also cooperates with a brake added between a rocker part and bogie to
Mars rover lift the sunken wheels separately. The strategies of wheel-step motion, wheel lifting, and
suspension folding/unfolding are practicable as validated by prototype tests. Experiments
show that the active suspension considerably enhances the rover’s ability to escape from
a dune or a crater. The proposed suspension is an important attempt in improving the
design of Chinese Mars rovers.
© 2018 Elsevier Ltd. All rights reserved.
1. Introduction
Mars is the most probed planet in the solar system because it is the most likely to contain extra-terrestrial life and it
contains rich minerals that are rare on Earth. The development of science and technology has allowed major space powers
to launch spacecraft and other vehicles to Mars. The space agencies of USA [1–4], Russia [5], China [6], Europe [7], India [8],
and Japan [9] plan to send their own probes/rovers to Mars. An increasing number of scientists have focused on the design
of mobile rover systems. Some have researched the existing Lunar and Martian rovers for many years and obtained several
configurations, including wheeled robots, wheel-step robots, and legged robots.
Wheeled robots have a higher efficiency than legged robots, but their mobility is relatively poor. Wheel-step robots can
use both wheeled and legged motion, providing good efficiency for travelling over flat ground and traversing on rugged
ground. Previous Mars rovers were typically wheeled rovers with increased mobility provided by specially designed sus-
pension and wheels. The USA have successfully landed four rovers on the Martian surface, but two rovers launched by the
Soviet Union unfortunately crash-landed. The structures of those rovers are described in some papers [10–13]. The four
∗
Corresponding author.
E-mail address: zhenliu.hit@gmail.com (Z. Liu).
https://doi.org/10.1016/j.mechmachtheory.2018.05.002
0094-114X/© 2018 Elsevier Ltd. All rights reserved.
126 J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149
Nomenclature
American Mars rovers (Sojourner, Opportunity, Spirit, and Curiosity) are all wheeled robots with a rocker-bogie suspension.
Their mobile systems lasted longer than expected, and they were able to obtain valuable information. The American Mars
rover homepage has released malfunction information and driving difficulties encountered during operation. Spirit experi-
enced a deep sinkage twice and was eventually trapped in the Troy dunes. Opportunity also experienced a deep sinkage but
was eventually able to extricate itself from the sand trap. The rover carefully entered and left the Victoria Crater using a
considerable amount of time to find a safe route. A future rover must have a greater ability to traverse soft terrains, climb
steep slopes, and lift malfunctioning wheels.
The new suspension design should provide additional locomotion for the rover, delivering more traction than the max-
imum provided by wheel locomotion. The wheel–soil interaction of drive wheels is researched by some scientists. Bekker
presented a soil modulus model that was dependent on a sinkage exponent, and he considered the traction force mainly
produced by the soil in the entry contact area [14]. Wong and Reece improved Bekker’s model through an experimental
investigation by using two dimensionless parameters, and they considered the traction force produced in the exit area due
to wheel slippage [15]. Liang Ding et al. analysed the problem of longitudinal slip and skid of a wheel that occurs on a
slope using terramechanics [16] and studied the drive wheel’s performance during its motion in deformable soil [17]. The
terramechanics formulas can be revised and their errors can be reduced based on experiments. Considering the stress in
the exit area is a suitable method for increasing the calculation precision of terramechanics formulas. The exit contact angle
is approximately −0.2 times the entry contact angle. Senatore et al. proposed a normal stress and wheel–soil contact angle
based on experiments, and they considered stress in the exit area [18]. The integral of stress on the entry contact surface
is much greater than that on the exit contact surface. Shibly et al. proposed simplified equations of drawbar pull and com-
pared them with the original equations [19]. The shear stress in the low-cohesion soil is very low at the free surface. Guo
et al. proposed a normal stress distribution based on experiments in which the integral of stress on the exit contact sur-
face is ignored [20]. Ignoring the exit area is suitable for simplification and the model error is acceptable. This study uses
integration and stress distribution to research the wheel–soil interaction, and the soil stress in the exit area is ignored. The
terramechanics of the wheel-legged locomotion is different from that of the wheel locomotion. This study uses single wheel
tests to analyse the differences.
In previous wheeled rover experiments, we found that a braked wheel has a higher adhesion force than that of a ro-
tating wheel. The new locomotion design could use this phenomenon to obtain enhanced mobility. The rover wheels are
divided into dragged wheels (similar as anchored in soil) and drive wheels. The rover suspension has a deformation ability
as the dragged wheel is fixed on the ground while the drive wheel is moving. This locomotion is categorised as wheel-step
locomotion. Some suspension structures of wheel-legged rovers are good references for this study. Grand et al. proposed the
kinematics of a wheeled-legged rover (Hylos) with deformable suspension. The rover had four legs, and each leg had two
links and two revolute joints [21]. The control of this rover was complex because it used eight joints to change its geometry.
Inotsume et al. proposed a rover with four wheels. Its suspension could change the roll angle of the rover based on a linear
motion pair [22]. The linear motion pair had worse reliability than the revolute motion pair as it was easily locked by dirt.
Tarokh et al. proposed the kinematics control of a rover (SRR) with deformable suspension having an active revolution joint
between the rocker and body [23]. The deformation scheme of SRR is a suitable scheme. However, the design continuity
requires that the new suspension should inherit the general structure of the previous suspension, and the changes should
be as small as possible. The previous Mars rovers and the first Chinese lunar rover ‘Yutu’, all use a rocker-bogie suspension.
The new rover in this study combines the configuration of the SRR rover and the structure of a normal rocker-bogie sus-
pension. The concept and preliminary design method of the suspension were first proposed by Zheng et al. [24]. This study
adds more contents related to the design details to make them more specific and clear.
The basic characteristics of the rocker-bogie suspension need to be carefully studied, and its design requirements should
be summarised. The primary work would consist of changing the suspension to a wheel-step motion. Heverly et al. (JPL)
compared the Mars Science Laboratory rover and presented their traversal performance: a sharp transition between a moder-
ate slippage on a 10° slope and a vehicle embedding on a 17° slope [10]. Chottiner studied how varying parameters affected
128 J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149
the traversal performance and obstacle-climbing ability [1]. Miller et al. (University of Oklahoma) investigated the obstacle-
climbing ability of the rocker-bogie and showed its advantages in reducing the pitch angle of the body [25]. Its mobility
is related to the geometric parameters of the suspension. Li et al. researched the mobility performance of the rocker-bogie
suspension and proposed an optimisation of its parameters [26]. Kim et al. studied a stair-climbing mobile robot with a
rocker-bogie suspension, and they proposed the relationship between the travelling ability and the geometric parameters of
the suspension [27]. Choi et al. proposed a mobile platform (RHyMo) based on the rocker-bogie suspension, and the platform
has advantages in moving on rugged terrain [28]. These scientists verified the good mobility of the rocker-bogie suspension
and proposed its maximum capacity. This study proposes the design requirements for a normal rocker-bogie suspension and
develops an active rocker-bogie suspension based on the research of those scientists.
The rocker-bogie suspension of previous American Mars rovers is deformable to reduce the overall volume, but the defor-
mation is one-time and needs a supplementary mechanism. Pivirotto (JPL) presented folding and unfolding in the Pathfinder
Micro Rover [29]. Harrington et al. (JPL) presented the design of the Mars Exploration Rover (MER), pointing out that the
challenge posed by the MER was that the suspension should have the flexibility to stow in the limited available space [11].
Thus, the rocker is divided into two parts at the rocker pivot. The joints between the two rocker sections and body are in-
dependent and normally driven by a spring. These joints will be locked before the rover travels. This study uses joints at the
rocker pivot to produce the wheel-step suspension. We call this an active rocker-bogie suspension. The difference between
the suspension in this study and previous suspensions is that the folding/unfolding process can be repeated. This is useful
for avoiding bottom contacts with rocks and increasing the stability of the rover on slopes.
The mechanism and mechanical theory behind this solution are much simpler than other options. The rover has good
wheel-step ability, effective wheel lifting ability, and autonomous suspension unfolding ability. This study includes five
works: 1) verifying that the mobility of wheel-step locomotion is greater than that of wheel locomotion and deciding how
to implement a dragged wheel, 2) presenting the strategy of suspension deformation during wheel-step locomotion and
wheel lifting, 3) presenting the requirements for both normal and active rocker-bogie suspension, 4) optimising the geomet-
ric parameters of the suspension that meets the requirements in the second work, and 5) obtaining the transmission ratio
of the two joints. The performance of this suspension is verified through prototype tests.
The American Mars rovers carried out patrols on various terrains and obtained a considerable amount of information. The
missions of the original Mars exploration rovers (Spirit and Opportunity) were planned to last for 90 days. However, Oppor-
tunity had worked for more than 13 years and travelled approximately 43 km, while Spirit worked for 6 years and travelled
7.7 km. They both experienced wheel sinkage and slip, losing traction many times. The suspension functions proposed in
this study are designed based on an analysis of the pictures obtained by these American rovers.
It is not possible to analyse the physical properties of the soil on the road ahead of a rover. When the soil parameters
cannot be predicted, the road would be dangerous. One danger is when soft and rigid terrains are intermingled. The rover
would have difficulty in moving when the wheel sinkage is deeper than the wheel radius. Another danger is when a very
soft terrain is shielded by duricrust. The image would provide the operator incorrect data because the collapsed area would
appear similar to other regions; thus, wheel sinkage could occur suddenly. Moreover, traversing through loose terrains for
an extended time is a challenge for the wheels: this could cause a significant increase in wheel failure rate.
The maximum mobility of a wheeled rover is limited by the soil parameters. The rover requires strengthened mobility
on loose terrain and steep slopes to extend its detection range. The wheels should have sufficient traction without large
slippage and sinkage. This study investigates a novel locomotion and proposes its mechanism foundation and terramechanics
principles.
Wheeled rover traction is related to wheel–soil interaction. The maximum mobility is determined by the properties of
normal stress and shear stress. The upper limit of the slope-climbing ability of a wheeled rover is predicted using a single
wheel test, and the traction loss caused by excessive sinkage is given by a terramechanics analysis.
Table 1
Physical and mechanical parameters of the Mars soil simulant.
Parameter ρ c n0 n1 kc kϕ ϕ K
(kg/m3 ) (kPa) (kPa/mn-1 ) (kPa/mn ) (°) (m)
Table 2
Geometric parameters and material of test wheel.
Rim diameter Rim width Radian radius Tooth height Material Weight Static load
The test wheel properties are given in Table 2. The wheel cavity is open and the wheel face has sufficient strength and
stiffness. The wheel structure is not cylindrical to reduce the steering torque. The wheel deformation can be ignored under
a load of 0–500 N.
The test wheel is a drive wheel, and its slippage is controlled by the testbed. The relationship between the pull coefficient
and slip ratio in the wheel–soil interaction tests is shown in Fig. 2. The initial wheel load (normal force) is 100 N. The pull
coefficient increases significantly when the slip ratio is less than 0.3, and it increases very slowly when the slip ratio is
between 0.3 and 0.6. In past experience, the pull coefficient was nearly constant when the slip ratio was greater than
0.6. The maximum pull coefficient of the wheel in the simulant soil is approximately 0.4. The maximum slope at which
a wheeled rover can climb is obtained using Eq. (1) (approximately 22°). The pull coefficient of a wheel with a load of
20 0–50 0 N and slip ratio of 0.05–0.6 was tested in the same manner as the wheel with the 100 N load. The pull coefficient
of wheels with the same slip ratio increased along with the normal pressure. The change rule of their pull coefficient was
130 J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149
similar to that of the wheel with the 100 N load. Thus, the soil parameters determine the maximum slope-climbing capacity
of the wheel.
F
α ≤ arctan DPd
= arctan (0.4 ) = 22◦ (1)
FN
Considering the wheel pressure distribution when the rover climbs up a slope, the downhill wheel takes a heavy load
and the uphill wheel takes a light load. The maximum slope that a wheeled rover can climb is less than 22° because the
wheels provide less total traction than the sum of the traction when they work independently with an average load. The
climbing ability of a wheeled rover with the same wheel size and same wheel load is tested, and its maximum climbing
angle is approximately 20°. In comparison, the angle of the steepest slope that Opportunity climbed is approximately 15–17°.
Fig. 4. Relationship between the drawbar pull and entry contact angle.
τ1 (θ ) = [c + σ1 (θ ) tan ϕ ] × 1 − e− j (θ )/K (6)
Because 1 − e− j (θ )/K 1, the shearing stress has an upper limit, as indicated in Eq. (7). The cohesiveness c of the Mars
soil is considered as 0.25.
τ1 (θ ) c + σ1 (θ ) tan ϕ (7)
The drawbar pull FDPd of the drive wheel is the integral of the soil stress p1x on the wheel–soil contact surface, and it
can be calculated using Eq. (8). It is equal to the parallel component fDP of the wheel load. From Fig. 2, we can observe that
it increases slowly once the slip ratio exceeds 0.3.
θ1 θ1
FDPd = fDP = br p1x (θ )dθ br σ1 (θ )(tan ϕ cos θ − sin θ )dθ (8)
0 0
Spirit and Opportunity lost their mobility when their wheels had excessive slip-sinkage. It is not possible for wheels to
rotate without slippage in wheeled rovers. Therefore, increasing traction by drive wheel has a limitation. This study proposes
the use of a dragged wheel to provide enough traction and a deformable suspension for the transfer of the drive wheel.
The dragged wheel is used to provide adhesion force to reduce slippage of the drive wheel. The initial sinkage of the
dragged wheel is caused by slip-sinkage of the drive wheel. The adhesion force of the dragged wheel is mainly dependent
on its sinkage and soil accumulation at the back of the wheel. The direction of the adhesion force of the dragged wheel is
opposite to its linear velocity v. The wheel load is borne by the back contact area because the wheel tends to separate from
the soil in the front contact area, as shown in Fig. 5.
The dragged wheels easily provide a greater adhesion force than a drive wheel with the same size if they have the same
sinkage. The slip-sinkage of the dragged wheel is weaker than that of the drive wheel when the same adhesion is required.
Thus, the rover can pass through a looser sand trap and climb a steeper slope.
The relationship between the adhesion force provided by a brake wheel and the drawbar pull of a drive wheel is also
verified by a wheel–soil interaction test based on the same test bed and simulant soil as in Section 2.2. The vertical load
on the test wheel is also 120 N, and the initial static sinkage is 0.03 m. The initial sinkage of the dragged wheel is made by
a stable sinkage of the drive wheel with a slip ratio of 0.1–0.9. The adhesion comparison between the dragged wheel and
drive wheel is shown in Fig. 6. The wheel rotation speed is 1.0 rpm, and the slip ratio is controlled by the test bed drag
132 J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149
motor. The drive wheel works for 92 s in order to obtain stable sinkage, and its drawbar pull is small. Then, the drive wheel
is switched to the dragged wheel with the same linear velocity as the drive wheel. The adhesion force of the dragged wheel
increases significantly and quickly in approximately 15 s. Then, the adhesion increases at a slower rate and approaches a
stable value. The incremental adhesion force between the dragged wheel and the drive wheel is approximately 100 N.
The results of wheels with a slip ratio of 0.2–0.9 are shown in Fig. 6. These wheels work for approximately 92 s in order
to obtain a stable sinkage. The linear velocity of the dragged wheel is also 1.0 rpm. The drawbar pull of the drive wheel
increases with the slip ratio, and the adhesion force of the dragged wheel increases with an initial sinkage at 92–105 s. The
stable adhesion force is also related to the initial sinkage, and the dragged wheel with a slip ratio of 0.9 can provide the
greatest adhesion force. The incremental adhesion force is approximately 150 N.
The height at the back of the wheel increases and the entry area soil is compacted when the wheel begins dragging.
Thus, the adhesion force increases quickly. After time passes, the soil friction angle stabilises the height at the back of the
wheel. Thus, the adhesion force increases slowly after 105 s. The traction coefficient = FDP /QW is used to reflect the wheel
adhesion capacity. FDP is equal to FDPd when the wheel is a drive wheel. FDP is equal to FDPb when the wheel is a brake
wheel. Its maximum value is approximately 0.4 when the wheel completely slips, and in dragged locomotion, it ranges from
0.87 to 1.67 once the soil accumulation is stable. The dragged wheel has more adhesion force than the drive wheel.
The dragged wheel and the drive wheel are combined to form a wheel-step locomotion based on a new suspension.
The core features of the suspension are the ability to change the distance between two wheels and switching the wheels
between dragging and driving. The locomotion principle of the wheel-step rover is shown in Fig. 7. The joint between link
l1 and the body rotates along with the joint between link l2 and the body. The distance between the dragged wheel and
drive wheel is changed, and the drive wheel moves forward. Next, the previous drive wheel becomes the dragged wheel
and vice versa. The dragged wheel provides a strong adhesion force, and the drive wheel not only provides drawbar pull
but also permits wheeled motion for body motion.
J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149 133
This model allows the rover to use fewer joints for both the wheel-step motion and the folding/unfolding suspension
process. Although the horizontal and vertical motions of the body are not independent, this coupling movement can be
acceptable compared to using more joints. The use of a linear pair is avoided to increase the reliability of aerospace products.
The suspension in Fig. 12 has a lower obstacle passing ability and a rougher ride than the normal rocker-bogie suspension
because it has no bogie. This study proposes a new rover based on a combination of this simple model and the rocker-bogie
suspension.
The suspension in this study inherits the basic structure of a normal rocker-bogie suspension, and it has an active defor-
mation ability. Thus, it is called an active rocker-bogie suspension. We demonstrate the configurations and functions of this
suspension in this section.
A normal rocker-bogie suspension has three passive joints, as shown in Fig. 8. These joints cause all six wheels to be in
contact with the ground. One of the joints is a couple joint between the two output shafts of the differential mechanism.
The other two are free revolute joints between rockers and bogies.
The rocker of a normal rocker-bogie suspension is divided into two parts at the rocker pivot. The wheel-step locomotion
is based on changing angles between the rocker parts and body by an angle-adjusting mechanism. The posture of the body
changes only slightly during suspension deformation. The deformation of the suspension and the condition of wheels during
wheel-step locomotion is shown in Fig. 9.
This scheme has several advantages. Using the least possible number of active joints and less driving numbers means
higher reliability. The joints are at the highest position of the suspension, and thus, they do not interfere with the links. The
joint driving mechanism is installed in the body, which is beneficial for heat preservation and dust prevention. In addition,
the suspension deformation during wheel-step locomotion can be used in the suspension folding/unfolding process.
The wheel-step locomotion can help extricate the rover from difficult situations. When the load on the joints is high and
the wheel sinks too deep, the wheel must be lifted up such that sand flows below the wheel. One of the three joints of the
134 J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149
Fig. 9. Pattern with two active wheel-step joints added on rocker pivot.
normal rocker-bogie suspension should be removed and one brake is added between the rocker and bogie. The rocker and
bogie are bonded when this brake is engaged. The deformation of the suspension and rotation of the middle wheel allows
any of the three wheels to be lifted independently. Moreover, both sides of the suspension can lift a wheel at the same
time.
The following sections describe the wheel-step locomotion, suspension folding/unfolding, and wheel lifting in detail.
Table 3
Wheel status during slope climbs.
Climbing mode Step Rear wheel Middle wheel Front wheel Incremental FDP
this study, is asymmetrical front to back. Moreover, the backward wheel-step motion has a better wheel load distribution.
However, the navigation camera is set at the front of the body, and the backward wheel-step can be only used as a backup
method.
Although the velocity of wheel-step locomotion is low, the rover has reliable mobility when climbing a steep slope and
traversing a soft terrain, which is of utmost importance.
Fig. 12. Single wheel is lifted while the rover travels on flat terrain.
Fig. 13. Two wheels are lifted while the rover travels on flat terrain.
Fig. 14. Pattern with one active wheel-step joint added on the rocker pivot.
More independent joints lead to more actuators and more complex control. If two joints are related by a mechanism,
only one actuator is required. The main problem is finding a mechanism that accomplishes the desired rover functions
described in Section 3. The transmission ratio between the two rocker joints is based on the geometric parameters of the
suspension.
Wheel-step locomotion may be realised using one revolute joint, as shown in Fig. 14. The change of the branch angle
between the front rocker and rear rocker (ϕ1 − ϕ0 ) leads to a change in the wheelbase between the front and middle
wheels. However, one joint will cause the body posture to change considerably. Another joint is required to reduce this
change. Therefore, we require a 1-in 2-out mechanism or a 2-in 2-out mechanism.
The angle-adjusting mechanism is added between the output shaft of the differential mechanism and rocker parts, as
shown in Fig. 15. It has two revolute joints and controls the angles between the rocker sections, between the front rocker
and output shaft of the differential mechanism, and between the rear rocker and output shaft of the differential mechanism,
as shown in Fig. 15(a). Because it is located in the body, the angle-adjusting mechanism has good heat preservation and is
dust proof, as shown in Fig. 15(b). The mechanism has a cylindrical structure to reduce the effect on instruments inside the
body.
The angle-adjusting mechanism is a 1-in 2-out mechanism because fewer motors and a simple transmission chain have
better reliability. The first output is the angle change between the front rocker and output shaft of the differential mecha-
nism. The second output is the angle change between the rear rocker and output shaft of the differential mechanism. These
two outputs are related because they use the same driving mechanism. The mechanism and its control would be very com-
J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149 137
Table 4
Planetary gear train with two outputs.
Part I II III
plex if the relationship between the two joints is nonlinear. Therefore, the relationship should be linear to reduce the design
difficulty.
A planetary gear train is a typical 1-in 2-out mechanism with a linear transmission. It has three configurations, as pre-
sented in Table 4. There is one basic configuration each for the sun gear, ring gear, and carrier.
The configurations given in Table 4 are shown in Fig. 16. The rotation directions of the two outputs are the same when
either the sun gear or ring gear is basic. The rotation directions of the two outputs are opposing when the carrier is basic.
The configuration in Fig. 16(c) is selected as the design scheme. In this configuration, the rear rocker has a different rotation
direction compared to the front rocker.
Fig. 17 shows a simple structural drawing of an applicable scheme of the active suspension. This angle-adjusting mecha-
nism consists of a planetary gear train driven by a reducer (R) and motor (M). The planetary carrier cris fixed on the output
shaft of the differential mechanism. The front rocker is fixed on the ring gear rg, and the rear rocker is fixed on the sun
gear sg. The levelness of the new rover is the same as in the normal rover because the differential mechanism is the same
as that of the normal rocker-bogie suspension. The angle-adjusting mechanism requires no power during normal wheeled
motion because there is a self-locking mechanism in the reducer.
The transmission ratio of the planetary gear train must be carefully selected because the body needs to remain parallel to
the platform during the suspension folding/unfolding process. In particular, the bottom of the rover would collide with the
platform if the transmission ratio is incorrect, as shown in Fig. 18. Consequently, the body would not have enough support,
and launch and transport safety would be decreased.
The transmission ratio is related to the geometric parameters of the suspension. The manufacturability of the planetary
gear train also limits the optimal transmission ratio. Thus, the geometric parameters should be carefully selected, and they
need to meet the requirements of the normal rocker-bogie suspension. The special requirements of the active rocker-bogie
suspension make it more difficult to select the parameters.
138 J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149
The five key geometric design points of the rocker-bogie suspension are O1 , O2 , W1 , W2 , and W3 . The mass centre C also
needs to be considered. The geometric parameters and structural parameters are shown in Fig. 19. They reflect the relative
position relationships among those key points. The main rocker branch angle is composed of α 1 and α 2 . The angle-adjusting
mechanism is used to change these two angles. The width of the rover is neglected.
J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149 139
Fig. 20. Relationship between middle wheel centre and centre of mass.
The design requirements for the normal rocker-bogie suspension and active rocker-bogie suspension are considered in
selecting these parameters. The parameters can be used to describe both the wheel-step motion and wheel lifting motion.
The requirements of the normal rocker-bogie suspension are considered to obtain a good wheel locomotion performance.
The structural parameters of the suspension are used to describe these requirements. Moreover, they can be converted to
the requirements of the geometric parameters.
(1) The distance between the body mass centre C and output shaft axle O1 affects the differential mechanism design. It
should be as short as possible: dc − dr ≈ 0. Furthermore, this is required for realising front and rear wheel lifts.
(2) When the rover traverses a flat ground, the pressure on one wheel should be similar to that on any other wheels. The
horizontal distance between O1 and W1 is twice that between O1 and O2 , and the horizontal distance between O2 and W2
is equal to that between O2 and W3 : dr = 2(db − dr ) and dbm = dbr .
(3) The rover should have similar forward and backward traverse abilities; as such, the wheel axle distance between the rear
and middle wheel is equal to that between the middle and front wheel: dfm = drm .
(4) The output shaft should be contained in the body for heat preservation, particularly when mechanisms are added on the
output shaft: hr + q < h1 + h2 and hr − q > h1 .
(5) The wheel diameter is designed based on an analysis of the obstacle size and terrain characteristics: d > d¯.
(6) The obstacle height that can be negotiated by the rover is not less than the diameter of the wheel, and the setting of
steering joint is considered: h1 ≥ d.
(7) The axle distance is larger than the diameter of the wheel: dfm > d, dmr > d.
(8) The total length of the rover L, maximum diameter of the output shaft q, ground clearance h1 , and body height h2 are
decided by the overall design formula: dfm + drm + 2r = L.
is linear. Fig. 21(a) can be changed to Fig. 21(b), where the vertical displacement of the end of the rear rocker is equal to
that of the end of the front rocker, as indicated in Eq. (10).
When the rotational speed of the two rotating joints is linear, it is easy to use a mechanism to realise the transmission
between them. The Pearson correlation coefficient [30] of ϕ m and δ r is represented symbolically by p. Its value is 1 when
ϕ m and δ r have a positive linear correlation, and −1 when ϕ m and δ r have a negative linear correlation. The angle change
between the front rocker and vertical during suspension folding/unfolding on horizontal ground is α 1 . ϕ m and δ r change
along with α 1 ,α 1 ∈ [a, b]. a, bare constants that describe the suspension deformation.
J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149 141
The range of α 1 during the total folding of suspension is [α 10 , α 11 ], as shown in Fig. 23. b is the difference between α 11
and α 10 , and it is calculated using Eq. (13).
h1
b = acos cos α10 − − α10 (13)
l1
The range of α 1 during body lifting is [α 10 , α 12 ], as shown in Fig. 24. a is the difference between α 12 and α 10 , and it is
calculated using Eq. (14). ho1 is the change in body height. If only the folding of suspension needs to be considered, a = 0.
ho1
a = acos + cos α10 − α10 (14)
l1
This study equally decentralises the variables ϕ m with 100 setpoints over the interval [minϕ m ,maxϕ m ]; n = 100is the
total number of discrete data. This study uses 1 − |p| as an optimisation objective in Eq. (15).
⎛ ⎞
n
(ϕmi − ϕ̄m ) δri − δ̄r
⎜ ⎟
aim : g(X ) = min (1 − |p| ) = min 1 −
⎜ ⎟
i=1
(15)
⎝
n n 2 ⎠
(ϕmi − ϕ̄m )2 · δri − δ̄r
i=1 i=1
4.5. Geometric parameters of suspension and transmission ratio of planetary gear train
Although suspension design is a complex problem, the core work is finding the optimal range of each parameter and
then selecting a group of parameters that meet the optimisation objective, as in Eq. (15). It is a typical single-objective
optimisation problem. The standard form of the optimisation algorithm can be formulated as follows:
⎧
⎪Minimize g(x )
⎪
⎪
⎪
⎪sub ject to :
⎪
⎨u j (x ) = 0, f or j = 1, 2, · · · , J
v j (x ) 0, f or j = 1, 2, · · · , J (16)
⎪
⎪A < w j (x ) < B, f or j = 1, 2, · · · , J
⎪
⎪
⎪
⎩X = (x1 , x2 , · · · , xn )
⎪
Designing the rocker-bogie suspension requires 6 independent design variables when l3 = l4 , α3 = α4 . The independent
design variables can be summed up as X = [l1 , l2 , l3 , α1 , α2 , α3 ] . ϕ m and δ r in Eq. (15) are functions of the variables of X.
The constraints of the parameters in Sections 4.3 and 4.4 are summarised as follows:
⎧
⎪l2 sin α2 = 2l1 sin α1
⎪
⎪l3 sin α3 = l2 sin α2
⎪
⎪
⎪
⎨l1 sin α1 + l2 sin α2 = 3l3 sin α3
st. h1 r, l3 cos α3 r (17)
⎪
⎪h1 + h2 > l1 cos α1 + r + q
⎪
⎪
⎪
⎩l1 cos α1 + r − q > h1
⎪
l1 sin α1 + l2 sin α2 + l3 sin α3 + 2r = L
This study uses the genetic algorithm toolbox in MATLABT to perform this optimisation. The initial value of X is selected
as [536, 411, 332, 68.5°, 90°, 65.5°] according to previous experience in designing a normal rocker-bogie suspension. A range
near the initial value is given to increase the calculation speed and ensure rationality of the results. The optimisation result is
feasible only when the transmission ratio meets both the principle of gear manufacturing and the principle of planetary gear
train design. The design results are as follows: the front rocker length l1 is 630.6 mm, the rear rocker length l2 is 313.4 mm,
142 J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149
the rear and front bogie lengths l3 ,l4 are 358.3 mm, and the branch angle of the bogie is 115.6°. In the normal mode, the
branch angle of the rocker is 156.4°, and in the fixed mode, the branch angle of the rocker is 222°. Fig. 25 shows the
relationship between ϕ m and δ r . The green line is a straight line, and the difference between the red line and green line
is small, which shows that the red line has good linear characteristics. Its linear slope is nearly −2.24/3.24 and is used to
design the planetary gear train.
The transmission ratio of the planetary gear train is calculated using Eq. (18). cr nsg is the rotational speed of the sun gear
with respect to the carrier, and cr nrg is the rotational speed of the ring gear with respect to the carrier.
cr
nsg
k= cr = − 2.24 (18)
nrg
5.1. Implementation
Fig. 26 shows the mechanical prototype of the active rocker-bogie suspension. The two planetary trains in the red rect-
angle are the core mechanisms discussed in this study. The rocker and bogie links are C-shaped plates to increase their
longitudinal stiffness. The brake is set at the position in the green rectangle. It is a toothed brake with more braking torque
than other types of brakes with the same size.
The prototype’s total mass is 220 kg. The ability of the rover to get out of trouble is much more important than loss of
time, and thus, the angle-adjusting mechanism is permitted to work slowly. The reducers include one worm reducer and
one harmonic reducer, as shown in Fig. 27. The total transport ratio is 3600:1, and the total transport efficiency is nearly
0.63. The motor drive torque is 0.54 N·m. The maximum permitted torque on the sun gear is 1222.4 N·m.
The planetary train is composed of two sun gears, four planetary gears, and one inner ring gear. The machining accuracy
and elasticity of the teeth ensure proper gear meshing, even if there are four planetary gears (four gears instead of three
evenly distributed gears). The gear parameters of the planetary gear train are given in Table 5.
The wheel-step locomotion test is a sloped climbing test. The wheeled rover loses its ability to traverse a slope when
the angle is more than 20°. Strategies to increase the slope-climbing ability should be researched before the test.
J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149 143
Table 5
Gear parameters of the planetary train.
m 2 2 2 2
z 29 18 65 29
Fig. 28. Sinkage when switching the wheel from driving to dragged.
Fig. 30. Normal force from wheel to ground when the suspension deforms on a slope.
Fig. 31. Comparison between the displacements of wheeled motion and wheel-step motion.
of mass and the wheel distance changes in the wheel-step motion, and the potential drawbar pull provided by each wheel
is changed. The centres of mass of the suspension parts are labelled as body mc, front rocker mc, rear rocker mc, and bogie
mc, as shown in Fig. 29.
When the suspension is folded, the rover sits down and the normal force of the wheel at the upper slope increases, as
shown in Fig. 30. For climbing a slope, the suspension folding from the normal wheeled mode is better than unfolding from
the normal wheeled mode.
5.2.3. Slope climbing comparison test between wheeled motion and wheel-step motion
For the wheel-step motion test, the rover wheels are prone to slippage when the rover climbs up a slope with soft
terrain. The slip ratio is more than 0.9 when the rover climbs a steep slope using wheeled motion. The slip ratio is 0.6
when the rover climbs the same slope using wheel-step motion. The maximum loose slope-climbing ability of the wheeled
rover is 20°. The maximum loose slope-climbing ability of the wheel-step rover is 25°, which is 25% more than that of the
wheeled rover, and it has the capacity to climb even steeper slopes. When the rover climbs a 20° slope, the rover in wheel-
J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149 145
Fig. 32. Comparison between treads of wheeled motion and wheel-step motion.
Fig. 33. Wheels are lifted and the rover travels on soft ground.
step motion travels farther than the rover in wheeled motion, as shown in Fig. 31. Thus, the wheel-step motion enables the
drive wheels to escape from sinkage and not to sink deeply.
The tread of the wheeled motion on a 20° slope is shown in Fig. 32(a). The wheel rotates for a long period to have the
same tread length as that of the wheel-step motion. The temperature of the driver is near the thermal protection limit. The
tread of the wheel-step motion on a 20° slope is shown in Fig. 32(b). The distance between both neighbouring treads in
the wheel-step motion is greater than that in the wheeled motion. Thus, the displacement loss of the wheel in wheel-step
motion is less than that in wheeled motion. The wheel-step motion has a greater climbing ability and stronger soft terrain
passing ability than the wheeled motion.
When the wheel sinkage is considerably deep, the sunken wheels must be lifted to decrease the sinkage prior to the
wheel-step motion. The strategy for the motion of a Mars rover is as follows: first, try the normal wheeled motion; second,
try the wheel-step motion; and third, try the wheel lift before the wheel-step motion.
The differential mechanism makes the deformation on both sides of the suspension independent. Lifting two wheels is
more difficult than lifting a single wheel because the stability of the rover is weaker. η1 is the rotational displacement of
the rear main rocker. The pitch angle of the body changes by approximately 1/3 η1 when both rear wheels are lifted, and
1/6 η1 when only one rear wheel is lifted. Fig. 33 shows the lifting of two wheels.
When the suspension deforms on a platform, the wheel rotation is controlled based on position because the friction
between wheels and the platform can be set according to the requirement. In other words, it is better than force control,
which can easily cause the system to diverge and produce unnecessary wheel slipping and vibration.
Fig. 36. Change in pitch angle and ground clearance when the rover stands up from flat ground.
Fig. 39. Folding rate of unfolding mode with respect to normal mode.
148 J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149
6. Conclusion
A dragged wheel has a greater traction force than a drive wheel with the same load. This is important for designing a
wheel-step rover that has a greater traction force than a wheeled rover. The wheel-step locomotion is produced by changing
the distance between the middle and front wheel. The angle-adjusting mechanism is used to change two joints between the
rocker sections and the body, which makes the wheelbase change. The core of the angle-adjusting mechanism is a plan-
etary gear train in which the relationship between the two joints is linear. This linear relationship is designed based on
the optimisation of geometric parameters. The planetary gear train is specifically selected to ensure that the body is par-
allel to the platform during suspension folding/unfolding. Each side of the suspension requires one degree of freedom to
be restricted by a brake when the rover lifts a wheel. The resistant torque from the middle wheel to the bogie works in
lifting the wheel during suspension deformation. The wheel-step locomotion provides the rover an excellent soft terrain
passing ability and slope-climbing ability. A malfunctioning wheel can also be lifted to allow the rover to continue travel-
ling. Suspension deformation, similar to wheel-step locomotion, is also used to reduce the rover volume in transport. The
advantage of this design is that the angle-adjusting mechanism is not only used in wheel-step locomotion, but also in wheel
lift procedures. The mechanism has a high usage rate, which is better than the previous one-time use assistant mechanism.
This design improves the performance of the mobile system at an acceptable cost. It is a good choice for future research
focusing on Chinese-led Mars missions. Future work focus on the coupling simulation of the rover based on dynamics and
terramechanics, and a control strategy of wheel-legged locomotion will be proposed.
Acknowledgments
This work was supported by the National Natural Science Foundation of China (51275106 and 51405109), the ‘111’ Project
(B07018), the Program for New Century Excellent Talents in University (NCET-10-0055), the Chinese State Key Laboratory of
Robotics and System Foundation (HIT.KLOF.2010057), and the Self-Planned Task (No. SKLRS201501B) of State Key Laboratory
of Robotics and System (HIT).
Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.mechmachtheory.
2018.05.002.
References
[1] J.E. Chottiner, Simulation of a Six Wheeled Martian Rover Called the Rocker Bogie Electronic Thesis or Dissertation, Graduate School of The Ohio State
University, 1992.
[2] M. Tarokh, G. McDermott, S. Hayati, Kinematic modelling of a high mobility Mars rover, International Conference on Robotics & Automation, 1999
May.
[3] G.A. Carr, L. Jones, V. Moreno, Mars Laboratory (MSL) Power System Architecture. 10th International Energy Conversion Engineering Conference, 2012
July 30 - August 1.
[4] D. Brown, S. Cole, G. Webster, The Mars science laboratory landing, World Neurosurg. 79 (2013) 223–242.
[5] W. Ji, Z. Guang-wu, Z. Hua, Scientific objectives of China-Russia joint Mars exploration program YH-1, Chin. Astron. Astrophys. 34 (2010) 163–173.
[6] Y. Peijian, S. Zezhou, R. Wei, Mission overview and key technologies of the first Mars probe of China, Sci. Chin. 60 (2017) 649–657.
[7] N. Patel, R. Slade, J. Clemmet, The ExoMars rover locomotion subsystem, J. Terramech. 47 (2010) 227–242.
[8] V. Sundararajan, Mangalyaan-Overview and Technical Architecture of India’s First Interplanetary Mission to Mars, American Institute of Aeronautics
and Astronautics Space Forum, San Diego, CA, 2013 September10-12.
[9] K. Matsumoto, N. Kamimori, Y. Takizawa, Japanese lunar exploration long term plan, Acta Astronautica 59 (2006) 68–76.
[10] M. Heverly, J. Matthews, J. Lin, Traverse performance characterization for the Mars science laboratory rover, J. Field Rob. 30 (2013) 835–846.
[11] B.D. Harrington, C. Voorhees, The challenges of designing the Rocker-Bogie suspension for the Mars exploration rover, 37th Aerospace Mechanisms
Symposium, Johnson Space Center, 2004 May.
[12] D. Shirley, J. Matijevic, Mars pathfinder microrover, Autonom. Rob. 2 (1995) 283–289.
[13] A. Kemurdjian, V. Gromov, V. Mishkinyuk, Small Marsokhod Configuration, International Conference on Robotics and Automation, Nico, France, 1992
May.
[14] G. Bekker, Theory of land locomotion (mechanism of vehicle mobility), University of Michigan press, Ann Arbor, 1956.
[15] J. Wong, and A. R. Reece. Prediction of rigid wheel performance based on analysis of soil-wheel stresses. Part I Performance of driven rigid wheels, 4
(1967) 81–98.
[16] L. Ding, H.B. Gao, Z.Q. Deng, Longitudinal slip versus skid of planetary rovers’ wheels traversing on deformable slopes, IEEE/RSJ International Confer-
ence on Intelligent Robots and Systems, 2013 November 3–7.
[17] L. Ding, H.B. Gao, Z.Q. Deng, Experimental study and analysis on drive wheels’ performance for planetary exploration rovers moving in deformable
soil, J. Terramech. 48 (2011) 27–45.
[18] C. Senatore, K. Iagnemma, Analysis of stress distributions under light weight wheeled vehicles, J. Terramech. 51 (2014) 1–17.
[19] H. Shibly, K. Iagnemma, S. Dubowsky, An equivalent soil mechanics formulation for rigid wheels in deformable terrain, with application to planetary
exploration rovers, J. Terramech. 42 (2005) 1–13.
[20] J. Guo, H. Gao, et al., Linear normal stress under a wheel in skid for wheeled mobile robots running on sandy terrain, J. Terramech. 70 (2017) 49–57.
[21] C. Grand, F. Benamar, et al., Motion kinematics analysis of wheeled-legged rover over 3D surface with posture adaptation, Mech. Mach. Theory 45
(2010) 477–495.
[22] H. Inotsume, M. Sutoh, K. Nagaoka, Modelling, analysis, and control of actively reconfigurable planetary rover for traversing slopes covered with loose
soil, J. Field Rob. 30 (2013) 875–896.
[23] M. Tarokh, H.D. Ho, Kinematics control and balancing of articulated rovers with active suspension, in: 28th International Conference on Computer
Applications in Industry and engineering, San Diego, United states, 2015, pp. 275–280. October.
J. Zheng et al. / Mechanism and Machine Theory 128 (2018) 125–149 149
[24] J. Zheng, Z. Liu, H. Gao, et al. A novel active deform and wheel-legged suspension of Mars rover. IEEE International Conference on Robotics and
Biomimetics. Qingdao, China, December 3-7 (2017).
[25] D.P. Miller, T.-L. Lee, High-speed traversal of rough terrain using a rocker-bogie mobility system, 5th International Conference and Exposition on
Robotics for Challenging Situations and Environments, 2002 June.
[26] S. Li, H. Gao, Z. Deng, Mobility performance evaluation of lunar rover and optimization of rocker-bogie suspension parameters, Syst. Control Aerosp.
Astronaut. 2008 (2008) 10–12 December.
[27] D. Kim, H. Hong, S. Kim H, et al., Optimal design and kinetic analysis of a stair-climbing mobile robot with rocker-bogie mechanism, Mech. Mach.
Theory 50 (2012) 90–108.
[28] D. Choi, Y. Kim, S. Jung, et al., A new mobile platform (RHyMo) for smooth movement on rugged terrain, IEEE/ASME Trans. Mechatron. 21 (2016)
1303–1314.
[29] D.S. Pivirotto, MESUR pathfinder micro rover flight experiment: a status report, Mars V Conference, 1993 May 26-29.
[30] K. Pearson, Notes on the history of correlation. Biometrika 13.1, (1920) 25–45.