IROS2019国际学术会议论文集 0309

Abstract Recently diverse research has actively been conducted to control the posture of jumping robots using an inertial tail mechanism However the inertial tail mechanism has a high probability of collision with obstacles In this study a momentum wheel mechanism is proposed to achieve the same attitude control performance while reducing the volume occupied by the inertial tail mechanism To verify the performance of the momentum wheel mechanism we proposed a jumping robot with a momentum wheel mechanism and performed a dynamic analysis simulation and experiments on a jumping robot with a momentum wheel mechanism In addition it has been demonstrated that the momentum wheel mechanism can contribute to control of the body angle of the jumping robot As a result the momentum wheel mechanism can enhance the stability of the jumping robot more than the tail mechanism and the momentum wheel mechanism contributes to the attitude control of the body angle which allows the jumping robot to perform continuous jumping I INTRODUCTION Reconnaissance robots should be able to overcome various kinds of obstacles stably and quickly to improve the efficiency of exploration 1 2 Jumping robots are suitable for reconnaissance robots because the jumping robot can quickly overcome obstacles However the jumping robot has a disadvantage that it cannot perform reconnaissance missions stably because the speed is too fast To overcome the various obstacles while enhancing the stability of the jumping robot it is necessary to design a leg model that has a stable trajectory even at high speed and requires an attitude control mechanism for controlling the posture of the jumping robot 3 6 Furthermore the jumping robot should be able to adjust the jumping height and jumping angle to overcome various height obstacles Recently studies on jumping robot leg design and the posture control mechanism have been actively carried out to increase the stability of jumping robots In 2009 a miniature jumping robot derived inspiration from the locust 7 They use a linkage structure to make the leg using one motor Also they use a circular structure to jump again after the robot drops on the ground However this robot has the drawback that it cannot control the jump angle and jump height Also since the volume of the balance control mechanism is large this robot cannot control its posture in narrow terrain In 2013 Jun Myeong Jin Kim is with the Department of Robotics Engineering Daegu Gyeongbuk Institute of Science and Technology DGIST Daegu 42988 South Korea e mail hambaf002 dgist ac kr Dongwon Yun is with the Department of Robotics Engineering Daegu Gyeongbuk Institute of Science and Technology DGIST Daegu 42988 South Korea corresponding author to provide phone 82 53 785 6219 fax 82 53 785 609 e mail mech dgist ac kr Zhang et al realized bio inspired jumping 8 This robot has a self righting function and the robot can adjust the jumping angle using an inertial tail Also the leg model is a 4 bar linkage structure and the robot can change the jumping angle adjusting the position of the inertial tail before jumping However the change in jumping angle is only 7 degrees and the time spent preparing for the next jump is 30 seconds As a result this robot is not suitable for reconnaissance robots which must have fast response capabilities In 2017 a research group created a jumping robot called Salto 1P that mimicked the Galago 9 They analyzed the leg angle and leg length of the Galago and made the leg of the robot using a linkage structure In addition this robot using an inertia tail and two thrusts to control its balance in the air to enable fast continuous jumping Using the balance control mechanism this robot can change the jumping angle from 40 degrees to 90 degrees Unlike the miniature jumping robot and the bio inspired robot Salto 1P can change the jumping angle and jumping height significantly because the jumping mechanism is more powerful than other robots 10 The jumping mechanism of Salto 1P is shown in Fig 1 The green circle indicates the start point of jumping and the black line indicates the jump trajectory When the Salto 1P first starts jumping the robot jumps vertically and controls the body angle using an inertial tail and two thrusters in the air to change the jumping angle of the next jump The reason the robot performs vertical jumping when first jumping is that vertical jumping allows it to reach the maximum jumping height and earn the time to control the balance At the second jump the jumping angle is changed depending on the body angle change In the third jump the robot can change the jumping angle by controlling the body angle without vertical jumping Figure 1 The jumping trajectory of Salto 1P jumping robot Salto 1P has high jumping performance due to excellent leg design the attitude control mechanism and jumping mechanism However despite the excellent jump performance the inertial tails and thrusters of Salto 1P have a pole shape which has the disadvantage of increasing the probability of collision with obstacles To resolve this problem it is necessary to consider a balance control Guinea fowl Jumping Robot with Balance Control Mechanism Modeling simulation and experiment results Myeongjin Kim and Dongwon Yun Member IEEE 2019 IEEE RSJ International Conference on Intelligent Robots and Systems IROS Macau China November 4 8 2019 978 1 7281 4003 2 19 31 00 2019 IEEE779 mechanism with the same control performance while reducing the volume occupied by the inertial tail and thrusters Furthermore for stable vertical jumping it is necessary to design a leg that makes a vertical trajectory while bearing the weight of the equipment attached to the reconnaissance robot In this paper we create a leg of a jumping robot that can make a vertical jump by mimicking the Guinea fowl model The reason for imitating the Guinea fowl is that it can withstand a heavy body weight of 1 3 kg and it can make the vertical jump stably and fast These advantages make it possible to perform vertical jumping stably even if the jumping robot has exploration equipment Next the momentum wheel mechanism was used to achieve the same attitude control performance while reducing the volume of the inertial tail The reason is that the momentum wheel can produce a large control output despite a small control input and the size of the momentum wheel is lower than the inertial tail mechanism The contents of the paper are organized as follows In Chapter 2 we designed the linkage structure of the leg and power transfer structure mimicking the Guinea fowl model In addition we made a prototype using the leg and power transfer structure We next simulated the Guinea fowl jumping robot to check the jumping performance Chapter 3 is introduces a fundamental study on balance control of the jumping robot using the momentum wheel and we made the dynamic model of the momentum wheel mechanism Also through the simulation and experimental results we proved that the momentum wheel mechanism contributes to the jumping robot s balance control Chapter 4 will cover the conclusion and future work II DESIGN OF GUINEA FOWL JUMPING ROBOT For the jumping robot to perform fast continuous jumping it is necessary to be capable of stable vertical jumping To create stable vertical jumping it is essential to mimic animals that have durable leg design and high vertical jumping performance The Guinea fowl can perform fast vertical jumping because of its leg design To accumulate jumping energy the Guinea fowl requires only 120ms For this reason Guinea fowl can perform rapid jumping Furthermore the Guinea fowl s legs are designed to allow stable vertical jumps at high speeds To realize these features we consider the joint angle of the leg over time and the ratio of leg length In Section A we design a leg model that has a linkage structure using the Guinea fowl s leg design Moreover we simulate the trajectory of the jumping robot s leg using the LINKAGE tool Section B will cover the calculation of jumping height and the design of the force transfer structure Section C presents a jumping simulation for a Guinea fowl jumping robot that combines the legs and power transfer structures A Design of leg model of Guinea fowl jumping robot The leg structure of the Guinea fowl help to withstand its body weight of 1 3 kg and can make vertical jumps 11 13 The high jumping performance of the Guinea fowl is due to the bone structure and the joint angle of the legs The leg structure of the Guinea fowl consists of a femur tibiotarsus tarsometatarsal and digits 11 12 As described in the reference 12 the ratio of bone length is 1 1 5 0 8 0 75 The ratio of the femur is 1 and the ratio of tibiotarsus is 1 5 the ratio of the tarsometatarsal is 0 8 and finally the proportion of a digit is 0 75 Next we considered the change of the joint angles of the legs over time The joint angle of the Guinea fowl is composed of the hip angle knee angle ankle angle and toe angle At 0 seconds the Guinea fowl has the maximum jumping energy and the hip angle knee angle ankle angle and toe angle become 30 degrees 54 degrees 45 degrees and 160 degrees respectively At this time the Guinea fowl has the maximum jumping energy and this stage is called the pre take off stage PRT stage After 0 seconds the Guinea fowl starts to stretch its legs Therefore the angle of the hip knee and ankle begins to increase and the angle of the toe starts to decrease When the Guinea fowl reaches the toe off state 120ms it begins to stretch its limbs as much as possible At this time the hip angle knee angle ankle angle and toe angle are 98 degrees 97 degrees 154 degrees and 173 degrees respectively This state is the post take off stage POT stage 11 a b c Figure 2 Modeling and simulation of linkage structure for Guinea fowl jumping robot We designed the linkage structure of the jumping robot based on the bone structure and the joint angles of the legs varying over time The design of the jumping robot s leg model using the linkage structure is shown in Fig 2 The blue line is the trajectory of the movement of the legs To verify the jumping trajectory of the Guinea fowl linkage model the revolute joint that connected to the femur must perform up and down motion The mechanism for moving the revolute joint is 780 discussed in Section B Fig 2 a is the PRT stage of the Guinea fowl model and the POT stage is Fig 2 b When the linkage model of the robot changes from the PRT stage to the POT stage the legs of the robot are unfolded to the maximum and the robot emits the jumping energy vertically Next we considered the relationship between the bone length and jumping angle as shown in Fig 2 c We analyzed how the jumping angle changes when the length of each link is increased by 1 mm When there is no change in the length of each link the jumping angle is 91 591 degrees In the case of the femur the jumping angle changed from 91 59 to 115 28 degrees as the link length changed and the jumping angle increased from 91 59 to 108 659 degrees when the length of the tibiotarsus was increased However in the case of the tarsometatarsal the jumping angle decreased from 91 591 to 83 89 as the link length was increased Through Fig 2 we can confirm that the designed linkage model has vertical jumping performance and we can adjust the jumping angle by adjusting the length of each link B Design of Power transfer structure The power transfer mechanism is an essential part of the jumping robot to move the jumping robot s leg This mechanism must be able to accumulate or release the elastic energy to move the legs In this section we used two linear springs to make the power transfer structure The force transfer structure by using the two linear springs is shown in Fig 3 Fig 3 a shows a cross section of the power transfer structure As shown in this figure the force transmission structure consists of a DC motor 2 1 reduction gear cam and two linear springs The height of the robot is 143 5mm at the PRT stage and 286 4mm at the POT stage In addition it can be confirmed that the slider is installed so that the linear spring can be compressed 21mm only vertically When the cam passes the critical point the two linear springs push the slider and release the elastic energy The operation of the force transfer structure is as follows First the DC motor moves the cam until the cam reaches the critical point When the cam passes the critical point the compressed linear spring releases For this reason the slider is raised and the linkage model moves Next we considered the jumping height to determine the spring constant of the linear spring The equation for the jumping height is given as Equation 1 In 1 is the number of springs is the spring constant is the total weight of the robot is the length of the compression of the linear spring and g is the acceleration of gravity 2 2 1 To implement the power transfer structure we used the two linear springs The number of springs N therefore is 2 Next to obtain the compression length of the linear spring and the total weight of the robot we used the I property function in Autodesk Inventor program We thus can know that the value of is 21 mm and is 150g Also we set the jumping height to 30 cm The reason for this is that the goal of the simulation for the prototype is to check whether the jump trajectory is vertical To determine the spring constant of the linear spring we substitute the parameters into 1 Finally we can find that the spring constant is 1 0 N mm a b Figure 3 Modeling of Guinea fowl jumping robot and the change of the spring energy over time Next for the force transfer structure to move smoothly the force of the cam must be larger than the spring force To calculate the force of the cam and spring we consider Equation 2 and Equation 3 Equation 2 obtains the force that can be produced by the cam The torque of the motor is 88 26Ncm the maximum diameter of the cam is 3 95cm and the reduction gear ratio is 2 2 3 Equation 3 represents the force of the spring The number of springs is 2 the spring constant is 1 0 N mm and the compression length of the linear spring is 21 mm Through 2 and 3 we can know that the force of the cam is 46N and the spring force is 42 N As a result the cam can push the two springs sufficiently In addition to verify the performance of the power transfer structure we considered the relationship between the angle of cam and the spring energy The relationship between the angle of cam and spring energy is graphically shown in Fig 3 b At 0 second the angle of the cam is 0 degree and the spring energy has a maximum value of 466 68Nmm When the cam passes the critical point the spring energy is released during 0 1 second and the angle of the cam after maximum emission becomes 56 6 degrees After 0 1 s the cam rotates counterclockwise and compresses the two linear springs to increase the spring energy The spring energy has a maximum value again at 0 62 seconds 781 C Jumping simulation Through the jumping simulation we wish to check whether the Guinea fowl jumping robot can vertical jump stably To perform the simulation we used the dynamic simulation function of Autodesk Inventor In Fig 4 the jumping robot is located on the jumping platform and the Guinea fowl jumping robot is maintaining the PRT stage To confirm the jump trajectory of the prototype the leg of the robot should change from the PRT stage to the POT stage a b Figure 4 Jumping simulation of Guinea fowl jumping robot Fig 4 a shows the process of the jumping motion At 0 second the robot has the maximum jumping energy as the PRT stage After 0 seconds the cam passes the critical point and the robot starts to jump At 0 13 seconds the robot reaches the POT stage At this time the leg is fully stretched and the foot is raised due to the reaction force The time taken from PRT to POT is 130ms which is different from the Guinea fowl model by 10ms The jumping robot reached a maximum jumping height at 0 4 seconds However at 0 13 seconds the robot tilts backward due to the center of mass of the robot For this reason the body angle of the jumping robot tilts up to 49 6 degrees at the maximum height At 0 64 seconds the robot falls to the ground The simulation results of the jumping height are shown in Fig 4 b Through the graph we can see that the maximum jumping height is 308 7mm and the robot falls to the ground at 0 64 seconds Through Fig 4 we confirmed that the jumping robot could perform a vertical jump However it can be seen that a balance control mechanism is required to implement the robot s landing motion for continuous jumping To achieve a stable landing using the balance control mechanism this mechanism should be used at 0 4 seconds to offset 49 6 degrees in the air and the robot should be positioned perpendicular to the ground Furthermore to prevent the robot from tilting at 0 13 seconds the balance control mechanism must be operated before the jump The details of the balance control mechanism will be discussed in chapter 3 III BALANCE CONTROL MECHANISM To land on the ground after the robot has jumped a balance control mechanism has to be installed in the robot 14 17 In the case of Salto 1P an inertia tail and two thrusts were attached to the robot to control the attitude of the robot in the air This method can control the balance well but the inertia tail has a high probability of colliding with surrounding obstacles To solve this problem we considered a balance control method using a momentum wheel instead of an inertia tail The momentum wheel has the advantage that it can control the posture more safely than the inertia tail when it hits an obstacle Section A will cover the design of the momentum wheel mechanism and we will carry out the momentum wheel simulation in Section B Section C gives a dynamic analysis of the momentum wheel model In Section D we conduct the experiment and compare the dynamic model and experiment results Section E covers a simulation of continu