IROS2019国际学术会议论文集 0195

Abstract Soft robots are a burgeoning archetype in robotics due to their ability to perform intricate movements easily and seamlessly They serve as an ideal concept for realistically emulating animal movements However the majority of soft robots today are unable to vary their motions due to the coupled interaction of the nature of their kinematics and structural composition We therefore created Carpie a robotic caterpillar adapted from a modular pneumatic actuator Carpie is designed to perform a variety of caterpillar movements by tuning its mechanical structure through physical reconfiguration The robot has a completely soft body that enables a variety of movements and contortions into different shapes making Carpie is perhaps the perfect example to showcase a soft robot s aptness for animal mimicry We analyzed the robot s control aspect Its capabilities were measured by performing gait velocity studies on three different configurations Each configuration was executed by applying various modules on Carpie s structure without fully refabricating the entire assembly We also analyzed the accompanying change in the robot s maximum height during gait Finally we demonstrate how physical reconfiguration was able to radically alter Carpie s movement allowing it to perform a turning motion I INTRODUCTION Conventional rigid body robots lack compliance and are rarely capable of showing the versatility and adaptability of natural organisms 1 4 While they can integrate compliant links or impedance control to their systems this inclusion complicates the operation of the robot This limitation is further highlighted when robotics is expanded beyond manufacturing and industrial automation into domains such as healthcare field exploration 5 and cooperative human assistance 6 where compliance and versatility are needed 7 Soft robotics is a quick emerging multidisciplinary field that combines material science mechatronics controls and Biomimetics 8 Soft robots in contrast to traditional rigid bodied robots are made out of soft material that have little or no non deforming links They fully utilize the unparalleled flexibility of their base materials to create continuum structures giving them the potential to exhibit unprecedented sensitivity and adaptability 1 Nevertheless the majority of current soft robots are laboratory prototypes which lack the refinement and reliability of market ready products and to fill this gap soft robots must undergo multiple enhancements from an engineering perspective 8 One of the primary areas that may contribute to this improvement is the advancement of soft actuation techniques Pouya Ahmadian is with the Department of Mechanical and Industrial Engineering at the University of Toronto UofT Canada email pouya ahmadian mail utoronto ca Rainier F Natividad and Chen Hua Yeow are with the Department of Biomedical Engineering at the National University of Singapore NUS Singapore e mail rfnatividad u nus edu bieych nus edu sg Elastomeric actuators powered pneumatically are of a special interest to soft robotics since they can be light low cost easily fabricated and capable of providing non linear motion with simple inputs 9 Meanwhile pneumatic actuators can be classified roughly according to their motion contracting actuators expanding actuators twisting actuators and bending actuators 10 Despite the variety in these actuator types most soft pneumatic actuators are limited to only one mode of motion motion path 10 It is possible to achieve complex modes of motion with these actuators by combining multiple actuators in synchronistic and antagonistic fashions However this increases design complexity and cost Furthermore once fabricated most of these actuators are restricted to their predicted operating range of motion 11 12 Subsequently the bending mode of the actuator should be determined and finalized before the fabrication of the robot The development of a new pneumatic actuator type that is capable of undergoing multiple bending modes after the robot s fabrication can further increase the versatility of soft robotics Such an actuator capable of active reconfiguration was developed by Natividad and Del Rosario 13 14 This actuator was composed of a non inflating flexible plastic spine and multiple inflatable fabric modules These replaceable modules offer the ability of physical reconfiguration for different modes of motion even after the fabrication of the spine In contrast to typical soft robots the actuator is able to execute motions that were unplanned during the initial design and fabrication stage To further investigate the dynamic response of this actuator type as well as to test its unique characteristics we created a soft caterpillar robot that functioned with a modular actuator This caterpillar robot named Carpie is patterned from the inching caterpillar 15 Fig 1 and is able to exhibit multiple gaits and stances with physical reconfiguration of the modular actuator This paper illustrates the operating principles of this robot as well as demonstrates some of its multiple modes of locomotion While multiple works have investigated the dynamics of soft robotic worms 16 18 Carpie s unique characteristics warrant their own investigation Carpie s overall behavior can be changed physically through varied combinations of module geometry which allows for the mechanical tuning of its operation We will investigate the different strides for the forward locomotion as well as discuss a new approach for the sideway turning of the robot Throughout the paper we will observe how the intelligent Rainier F Natividad and Chen Hua Yeow are with the Advanced Robotics Center at NUS Chen Hua Yeow is also with the Singapore Institute for Neurotechnology at NUS Carpie A soft mechanically reconfigurable worm robot Pouya Ahmadian Rainier F Natividad and Chen Hua Yeow 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 IEEE787 selection of the inflation modules enables the caterpillar to exhibit multiple modes of motion post fabrication II METHODS A Actuator Design Caterpillar Spine Similar to the actuator developed by Natividad and Del Rosario Carpie s actuator is composed of a single flexible spine that was 3D printed out of a Thermoplastic Polyurethane TPU variant Ninjatek Ninjaflex E 18MPa and a series of closely packed elastic inflatable modules fabricated out of TPU and Polyurethane E 350 MPa and inserted in the spine via a Polyethylene PET pneumatic connector 13 14 Shown in Fig 2 A are the spine and inflation modules that make up the actuator The caterpillar spine is subdivided by internal blockages into five segments with each having three module openings 15 maximum modules for the whole spine Similar to the actuator fabricated in 13 19 each module opening is 10mm apart with an additional 25mm of material added on both ends of the spine This brings the total length of the spine to 200mm Fig 2 B The segments each have their own independent controlled air supply that allows the application of varying pressures on distinct parts of the spine The modules attached to these segments are similarly designed to inflate into a spheroid and are replaceable without the need for any special tools The spine is supplied with pressurized air through an external tether and acts as the channel that distributes the flow to the modules which causes them to inflate Since these modules are closely packed on the spine the continued inflation causes them to exert force on their adjacent modules which restricts each other s free expansions This induces a moment on the spine and subsequently causes the bending of the flexible spine The spine goes back to its original extended position as the pressure is released from the spine and the modules are deflated This contraction and expansion of the spine is the basis for caterpillar s up and down motion inching motion B Vacuum Heads Caterpillar Legs While modular actuation facilitates the controlled up and down movement of the spine an additional mechanism is needed to translate this vertical motion of the spine to a forward movement In nature inching caterpillars achieve this type of locomotion by binding their ends to the ground at specific points during the contraction and expansion of their bodies In such caterpillars when the posterior end is moved forward the middle of the body is lifted up and the body of the caterpillar resembles an omega The caterpillar then releases the front grip and extends the body forward forming a cantilever 20 This allows the caterpillar to alternatively move its front and back ends forward as the body is going up and down Carpie attempts to emulate the movement pattern of an inching caterpillar by employing the same gripping and releasing inching gate for the forward movement Vacuum grippers were attached to the spine body to allow Carpie to grip the ground Vacuum grippers were employed since they provide a simple and delicate method to perform grasping 9 Carpie s vacuum legs vacuum grippers were modeled by incorporating a round flat vacuum cup in a rectangular base that s attached to the spine Both the vacuum leg and the attaching pin joint were also 3D printed out of Figure 1 The nnching gait The thoracic contraction of an inching caterapillar resulting in the bending of the body up photograph by Katja Schulz distributed under a CC BY 2 0 license cropped to fit A Vacuum Legs Figure 2 A Exploded view of the actuator and the vacuum legs The actuator is composed of the flexible spine and the inflatable modules The vacuum legs consist of two round vacuum cups and a base that attaches to the spine via a pin join B A schematic representation of an inflated module and spine The module diameter and spacing d are illustrated C The expansion of Carpie s inflation modules resulting in the bending of the spine Mimicking the inching motion of a caterpillar Carpie would first anchor its anterior tail as the spine is inflated pulling the posterior leg forward Next the vacuum suction is reversed as the spine extends forward sliding the anterior leg At this point a single stride is completed and the cycle beings again with another inflation state Flexible Spine C Inflation Modules B 788 Ninjaflex The vacuum legs can be seen connected to the spine on Fig 2 C Each vacuum gripper was connected to a commercially available vacuum ejector capable of applying vacuum at selected points in the stride cycle When Carpie is moving forward the front leg is fixed as the spine is bending front leg vacuum engaged As the spine is expanding the front leg is released and the back leg is fixed instead back vacuum engaged This leads to the alternating movement of the front and back legs that makes up the forward gait C Control System An open loop control system was utilized for Carpie s operation Each spine segment was attached to its own solenoid valve which in turn was connected to a 100kPa positive pressure line Energizing opening the valve pressurized the segments while de energizing it exposed the segments to the atmosphere Similarly the vacuum legs attached to the ends of the spine were connected to similar valves which were subsequently connected to a single vacuum ejector Opening these valves effectively anchors the affected tail The ejector was constantly supplied with 300kPa of positive pressure Valves were controlled by their respective transistor switches governed by a microcontroller Arduino Mega 2560 The schematic of the robot s control system is shown in Fig 3 III EXPERIMENTAL DESIGN A Inflation Selection Pattern To highlight Carpie s ability for multimodal operation through physical reconfiguration a straight inching experiment was designed with three different spheroidal module diameters 15mm maximum diameter 20mm maximum diameter and 25mm maximum diameter For this straight inching experiment only 9 modules were used the middle three segments and no modules were attached to the outer two segments 1 one on each side This format was chosen to prevent possible collision of the outer modules with the vacuum legs at high bending angles To induce gait each segment was independently activated and deactivated according to a pre defined pattern based on the movement of the inchworm 21 A global timing variable called interval timing t was defined to control the time duration at which the segment valves are energized or de energized To start a stride defined as the motion from the beginning of each inflation phase to the subsequent inflation phase the valve for the middle segment segment 3 was opened for the duration of one interval timing to initiate pre contraction 0 t 22 Spine segments 2 and 4 were then activated along with segment 3 to achieve full contraction for a duration of two interval timings t 3t Finally all the valves were closed to deflate the modules for a duration of 3t 3t 6t The vacuum legs were also opened and closed accordingly Each inching trail was repeated for 6 interval timings t 100ms 200ms 300ms 400ms 500ms and 600m A lower interval timing would mean that a single robot stride happens more quickly as less time is allocated for each inflation or deflation phase Alternatively a higher interval timing allocates more time for the inflation and deflation phases on each stride Shown below in Table I is the programmed forward inching pattern for this experiment as well as the duration of the phase of the stride as multiples of interval timings t Seven differently colored circular markers were attached to the robot which allowed for visual tracking of various points on the spine and the two legs Beginning from rest Carpie s control system was activated while a non moving camera recorded the resulting movement the camera was positioned such that it was able to capture 30cm of horizontal displacement of the posterior tail Each module size was tested using the patterns outlined in Table 1 Each module i e 15mm 20mm and 25mm was tested under 6 different pattern timings each data set consisted of three trials resulting in 18 data sets The footage was captured at 120fps and was processed using Tracker Phylets org IV RESULTS The marker at the center of the spine middle of segment 3 was tracked and its coordinates were recorded for all 18 data sets Fig 4 A shows the recorded horizontal position of this mid spine point as a function of time for the 15mm 20mm and 25mm module sizes Note that for the 25mm alternative the robot did not demonstrate forward locomotion at the t 100ms For this trial the spine remained in a fully contracted position during the first stride as the valves looped in the opening and closing states without being able to deflate the modules We observed that 100ms was not enough to Figure 2 A schematic diagram of Carpie s control system The positive pressure lines negative pressure lines and electronic logic lines are colored green blue and red respectively An industrial compressor provides positive pressure TABLE I MODULE INFLATION SELECTION Stride Phase Stride Time a Posterior Leg Segment 2 Segment 3 Segment 4 Anterior Leg Pre Contraction 0 t Released Depressurized Pressurized Depressurized Anchored Full Contraction t 3t Released Pressurized Pressurized Pressurized Anchored Expansion 3t 6t Anchored Depressurized Depressurized Depressurized Released b a 0 marks the beginning of each stride The total stride duration is 6t b Note that the timing shown are a simplification of the actual control program 789 allow for any meaningful inflation or deflation on this configuration due to the large volume of the 25mm modules Meanwhile representative screenshots of the forward moving experiments are found in Fig 5 Horizontal speeds were determined by applying a linear fit to the displacement time curves for all the data sets Fig 4 B summarizes these horizontal speed values for the three module categories and the 6 different interval timings The 15mm 20mm and 25mm modules achieved peak velocities of 0 41cm s 0 99cm s and 0 84cm s at interval timings of 200ms 600ms and 500ms respectively Note that the 100ms interval timing for the 25mm module had a velocity of zero since it did not yield any forward movement zero horizontal speed To further investigate the differences in inching gait among the different configurations we examined the maximum vertical displacement achieved by the mid spine point across each of the 18 trials Fig 4 C This metric provides information on how much the spine is being bent during locomotion as the spine is bending the height at the middle of the spine increases Note the large maximum height value for the 25mm module at 100ms interval timing despite the velocity of 0cm s for this trial In this trial the spine loops continuously in the first contraction position without the modules being able to deflate In general there exists an upward trend between inflation timing and maximum height since the actuator is allowed a longer duration to inflate Meanwhile the 15 mm module generally has a lower horizontal speed and height clearance as compared to the 20mm and 25mm module alternatives with the exception of 25mm at 100ms which had a speed of zero This shows that Carpie s forward locomotion is slower and has a lower walking profile when using this smaller 15mm module size On the other hand the 25mm module has the highest height clearances as compared to the other two alternatives This further shows that Carpie s gait for this bigger module size involves more bending of the spine at the contraction phase Therefore the data conclusively shows that faster gait speeds can be achieved by increasing the bending curvature of the spine This provides the robot with a straightforward method to modify gait speed bigger modules will result in faster walking As a consequence larger curvatures will then result in a higher gait profile While such an effect is negligible in open environments it limits Carpie s ability to trav