1 2 3 A novel load cell-supported research platform to measure vertical and 4 horizontal motion of a horse’s centre of mass during trailer transport 5 6 7 G Robert Colborne a,*, Liqiong Tang b, Brooke R Adams a, Brooke I Gordon a, Bridie E McCabe a, 8 Christopher B Riley a 9 10 a School of Veterinary Science, Massey University, Palmerston North 4410, New Zealand 11 b School of Engineering and Advanced Technology, Massey University, Palmerston North 4410, New 12 Zealand 13 14 15 16 Corresponding author: 17 18 G Robert Colborne 19 School of Veterinary Science 20 Massey University 21 Private Bag 11 222 22 Palmerston North 4410 23 New Zealand 24 Email address: G.R.Colborne@massey.ac.nz 25 Phone: +64 6 356 9099 Ext 85185 26 27 28 29 30 mailto:G.R.Colborne@massey.ac.nz Abstract 31 32 During transport, horses are subjected to acceleration in three dimensions, rapid braking, turning, 33 noise and other stressors. The animal’s ability to make postural corrections may be insufficient to 34 prevent injury or distress, and so knowledge of the compensatory motion patterns of the horse in the 35 trailer is a necessary precondition for smart design of transport systems. A custom two-horse trailer 36 was built for this project. It had a horse compartment 1.85 m wide by 3.95 m long, with adjustable 37 bulkheads and a centre divider separating the horses. The floor was instrumented with 24 shearbeam 38 load cells to measure the vertical load imposed by each horse and its horizontal motion. Two horses 39 were driven on a 56km trip on both rural and urban roads. Load data were collected at 100 Hz for the 40 58-minute trip and were filtered with a cut-off frequency of 5 Hz using a Butterworth low-pass filter 41 and then vertical acceleration computed. A pivot table counted sign reversals in the vertical 42 acceleration signal, and vertical displacement was calculated using the fundamental frequency of the 43 resulting acceleration data. Total vertical motion was calculated by making the negative 44 displacements absolute and summing these with the positive displacements, and vertical work done 45 was calculated by multiplying the force by the displacement measures. Horizontal motion was 46 calculated by averaging the transverse and cranio-caudal position of the centre of pressure every 47 second and adding the resultant displacements. Absolute vertical displacement of the two horses was 48 69.55 m and 97.56 m. In addition to the work done by standing, vertical work done in response to 49 vibration was 322.4 kJ and 443.2 kJ. Horizontal excursion was 227.1 m and 243.0 m. This is a first 50 effort to quantify the additional workload imposed on animals during transport, which will aid in the 51 design of smart transport vehicles that will minimise the stress to horses. 52 53 Key words: 54 55 Horse 56 Forces 57 Horse trailer 58 Load 59 Mechanical energy cost 60 Vibration 61 62 1. Introduction 63 Animals are frequently transported to provide us with food and for reasons related to work, 64 recreation and companionship [1]. In New Zealand alone, more than 43 million animals (including 65 horses) are transported at least once in their life, and many are transported more frequently [2]. Globally, 66 up to half of the 5 million horses transported to slaughter arrive injured [3]. These statistics are echoed 67 for cattle, sheep, and other transported livestock species [4]. Horses transported by road in non-68 commercial, low capacity trailers (floats) for recreational purposes face similar risks with more than 69 108 million journeys made in North America alone [5,6]. The potential for compromised behavioural 70 and physical welfare during transport is associated with driver behaviour, the inability of livestock to 71 cope with unanticipated vehicle movements, and a lack of welfare-oriented trailer or truck design [6-72 8]. During transport, animals are in a dynamic environment and are subjected to acceleration in three 73 dimensions, rapid braking, turning, noise and other stressors. These arise as a result of vehicle dynamics 74 that are influenced by driver skill and behaviour, road conditions, weather and other factors experienced 75 in transit [5]. The animal’s postural corrections may be insufficient to prevent injury or distress, either 76 because they exceed its physical abilities, or because of the varying behavioural (i.e. emotional) 77 responses of individual animals (and species) to these stressors. A key challenge is to determine which 78 of these animal responses (behavioural, kinematic, biomechanical or a combination thereof) to transport 79 dynamics are the most indicative and earliest indicators of adverse welfare, so that real-time feedback 80 can be implemented in smart vehicle design and initiate operator responses before injury or distress 81 occurs. Whereas various approaches have been used previously to evaluate energetics during equine 82 gait [9,10] there has been a lack of development of tools that facilitate the real-time measurement of 83 animal movement and the dynamic loads experienced by them during road transport. Earlier studies 84 have reported on gross movements of horses in transport using video. Untethered single horses in a 4-85 horse trailer left free to choose their own orientation spent significantly more time facing rearward than 86 in any other orientation during a 32km trip on country roads [11]. Total forward and rearward motion 87 was measured in horses confined to a compartment and travelling facing frontward or rearward during 88 a 14.4km trip around a track with turns and stops. Facing frontward, total cranio-caudal motion was 89 12.95m and facing rearward was 16.99m, although the range of movements across 12 horses was highly 90 variable and so not significant [12]. In both of the studies cited above, the investigators concluded that 91 individual horse effects were stronger than the effects due to orientation alone and neither provided a 92 basis on which to determine comparative workload of different transport orientations. 93 To facilitate a multidisciplinary approach to the problem of understanding the biomechanical 94 environment and its impacts upon animal behaviour and dynamic movement during transport, we 95 have assembled a team of researchers with expertise in livestock health, behaviour, transport welfare, 96 kinematics, mechatronics and data modelling. Anecdotally, horses are variably reported to prefer 97 transport facing forward, backward or on an angle, and a suitably designed research platform is a step 98 toward putting objective evidence to evaluate these beliefs. The objective of this paper is to describe a 99 custom-built horse trailer with load cells in the floor as a research platform for the estimation of 100 vertical and horizontal motions of the horse, and its mechanical vertical work output during road 101 transport based upon displacement of its centre of mass. 102 103 2. Materials and Methods 104 105 Ethics approval was granted by Massey University Animal Ethics Committee prior to 106 commencement of the project (MUAEC No. 17/97). 107 108 2.1 Trailer 109 110 A custom dual-axle two-horse trailer 1 measuring 2 m wide x 5 m long (excluding the drawbar) 111 was designed by the authors and constructed for the project. The inside dimensions of the horse 112 compartment measure 1.85 m wide x 3.95 m long, with adjustable bulkheads and a centre partition for 113 positioning a horse either side. The trailer has a progressive electric braking system with wireless 114 control that is automatically run from the towing vehicle, and an airbag suspension system with 115 onboard air compressor and 12V deep-cycle battery. There are two roof vents forward and a small 116 rectangular sliding window on each side at the rear for ventilation. There is no front window. The 117 trailer has an enclosed forward compartment for batteries and the measuring equipment. 118 The floor of the horse compartment consists of four independent sensing panels, each measuring 119 0.92 m x 1.15 m and constructed of 32 mm plywood with a 3 mm steel plate on its undersurface and 120 covered with an 8 mm high density non-skid rubber top surface. When positioned facing forward or 121 backward on either side of the centre partition, the front feet of the horse were on one panel, and the 122 hind feet on the other panel. Each floor panel is supported on six anodised aluminum shearbeam load 123 cells 2, one under each corner and one under the midpoint of each long edge. The wires from each 124 load cell are led to the front compartment of the trailer, through the floor and connected to six 3-125 channel National Instruments (NI) 9923 modules plugged into an NI c-RIO 9066 chassis (8-slots, 126 integrated 667MHz integrated dual-core controller) 3. A triaxial accelerometer 4 is interfaced to the NI 127 c-RIO chassis with a 3-channel NI 9230 module and attached to the front bulkhead of the trailer to 128 record accelerations of the trailer in the vertical, side-to-side and front-to-back directions during 129 motion. 130 131 2.2 Software 132 133 A data acquisition program was developed using NI LabView to collect the dynamic signals from 134 each load cell and from the accelerometer. Data were collected at 100 Hz directly to a laptop 135 computer, which was started and stopped manually at the beginning and end of each trip. Prior to 136 loading the horses, offset values were collected for 1 second to zero the transducer signals. The 137 horizontal (X= transverse, Y= fore-aft) location of each load cell and the load (kg) on it was used to 138 calculate the X-Y location of the centre of pressure of each horse between all four feet. This location 139 was recorded to disk at 100 Hz. The total vertical load (kg) generated by the horse on its two panels 140 was recorded from the 12 load cells and likewise stored to disk at 100 Hz. All data were stored in .csv 141 format so they could be post-processed in Microsoft Excel 5. 142 143 2.3 Horses 144 145 Two horses (both mares) belonging to the veterinary teaching herd at Massey University were 146 used. They were accustomed to forward-facing trailer transport but had not been recently moved by 147 trailer. Their masses were 462 kg and 471 kg. They were transported side by side facing forward in 148 the trailer and aside from being constrained within their compartments by the centre divider and their 149 fore and aft bulkheads, they had no other restraints. 150 151 2.4 Transport details 152 153 The trailer was pulled by a Toyota Hilux 2.8L diesel utility vehicle 6 driven by an experienced 154 driver. The round trip of 56 km included both rural and city roads, with some gentle hills, stoplights 155 and sharp turns. The maximum speed limits (80-100 km/h in the rural area and 50 km/h in the city) 156 were followed at all times, and the trip took about 58 minutes from start to end, which was the same 157 location. 158 159 2.5 Data processing 160 161 To establish the frequency range of the data recorded by the load cells using a rigid, passive load, a 162 four-legged steel bench weighing 223 kg was positioned on one floor panel and driven around a 163 shorter version of the normal trip (without horses) at the same speed. A Fast Fourier Transform (FFT) 164 was run on 4096 samples of the load data, collected at 100 Hz, to determine the frequency range 165 containing the most signal power. This was compared to the data from the accelerometer on the 166 forward bulkhead, also collected at 100 Hz at the same time. Figure 1 shows that the output from the 167 FFT on the load cell data under the steel bench indicated that most of the signal power was below 5 168 Hz. Subsequently, an FFT was run on the raw load cell data generated by a horse in transport, and that 169 likewise indicated that most of the signal power was below 5 Hz (Fig. 1). 170 Consequently, the raw total vertical load data (measured in kg) from the horse transport trial, 171 calculated as the sum of the individual loads from all the load cells under each panel, was put through 172 a low-pass Butterworth filter with a cut-off frequency of 5 Hz. This filtered load data was multiplied 173 by 𝑔 (acceleration due to gravity, 9.81 m/s2) to obtain force (𝐹) in Newtons and then vertical 174 acceleration (𝑎) was calculated from the following equation [Equation 1] using mass of the horse (𝑚) 175 and 𝑔: 176 𝑎 = (𝐹/𝑚) − 𝑔 [Equation 1] 177 This generated a long series of positive and negative acceleration values. These data were then put 178 through a pivot table in MS Excel to record the number of sign changes in the data, and to calculate 179 average acceleration values within each sign change. The fundamental frequency (f) of the data series 180 was determined as half the number of sign changes divided by the total number of samples in the 181 original data and the acceleration data was expressed as the alternating positive and negative average 182 values generated by the pivot table. 183 The instantaneous vertical displacement (Dvert) of the horse’s centre of mass was calculated [13] 184 from the following equation [Equation 2] using the fundamental frequency (f) previously calculated 185 and the acceleration values (𝑎) calculated in Equation 1: 186 𝐷𝑣𝑒𝑟𝑡 = (2𝜋𝑓)2 × 𝑎 [Equation 2] 187 This equation generated a series of positive and negative vertical displacements, the overall sum of 188 which was close to 0. The negative values were converted to positive, and then all values were 189 summed to represent the total vertical motion of the horse’s centre of mass during the transport trip. 190 To calculate the vertical work done by the horse, it was necessary to re-calculate vertical force 191 from the average acceleration values yielded by the pivot table. Vertical force (𝐹) was calculated 192 according to the Equation 3: 193 𝐹 = 𝑚 × (𝑎 + 𝑔) [Equation 3] 194 Work was then calculated by multiplying 𝐹 by the absolute vertical displacement at each interval 195 in the pivot table and summing all the values to generate a total value. 196 The raw X (transverse) and Y (forward-backward) positions of the centre of pressure under the 197 horse, collected at 100 Hz, were likewise filtered using a Butterworth filter at a cut-off frequency of 5 198 Hz. Then, the average of every 100 samples was calculated, and the resultant horizontal displacement 199 of the centre of pressure calculated every second using Pythagorean theorem, generating a map of the 200 horizontal excursion of the centre of pressure during the trip. The resultant X-Y displacement from 201 one second to the next was summed for a measure of total horizontal excursion during the trailer trip. 202 203 3. Results 204 205 The FFT on the combined signals from the six load cells under one panel bearing the steel bench 206 were a mix of true load plus higher frequency vibrations from the road surface. Most of the power in 207 the signal was contained below 5 Hz (Fig. 1a) and peak magnitude was at about 2 - 2.5 Hz. The FFT 208 on the accelerometer signal (Fig. 1b) indicates a broader range of frequencies with a less clear 209 dominant power, and reflects the vibration measured by the accelerometer due to the road surface. 210 The FFT from the load cells bearing one of the horses in its compartment (Fig. 1c) likewise indicates 211 that most of the signal was below 5 Hz and the peak amplitude was below 2Hz. At the sampling 212 frequency of 100 Hz, the horse transport trip generated a data stream of 346,916 samples (57.8 213 minutes). 214 After putting the acceleration data through the pivot table, and counting the sign reversals, the 215 fundamental frequency of the vertical load data was determined to be 1.94 Hz for Horse 1 (in the left 216 compartment) and this is likewise evident in the peak of the Fourier transform from that same horse 217 (Figure 1c). Figure 2 illustrates the vertical motion of the horse between reversals in acceleration. The 218 total (absolute) vertical displacement of the horse’s centre of mass was calculated to be 69.55 m over 219 the course of the trip. The fundamental frequency of the data generated by Horse 2 (in the right 220 compartment) was 1.65 Hz and its absolute vertical displacement was calculated to be 97.56 m. The 221 vertical work done was 322.4 kJ for Horse 1, and 443.2 kJ for Horse 2 during the trip, reflecting the 222 difference measured in total vertical displacement between the two horses. 223 The horizontal excursion of the centre of pressure was calculated to be 227.1 m for Horse 1, and 224 243.0 m for Horse 2. Figure 3 illustrates the excursion for Horse 1, and excluding extreme outliers, 225 the range of excursion transversely was about 0.32 m and cranio-caudally about 0.30 m. 226 227 4. Discussion 228 229 According to the Fourier transforms illustrated in Figure 1, the majority of the signal power from 230 the load transducers was below 5 Hz for both the trial using the weighted steel bench, and the trial 231 with the two horses, and this justifies the cut-off frequency of 5 Hz in the low-pass Butterworth filter 232 used to filter the data. The damping effect of the compliant limbs of the horse resulted in the peak 233 amplitude of the FFT frequency spectrum being slightly lower than the non-compliant steel bench, 234 and for the reduction of some of the higher frequency noise. Previously collected video recordings of 235 the limbs of the horse in a pilot study indicate substantial high-frequency joint motion as the limbs 236 attenuate vibrations caused by the road surface (unpublished data). 237 Calculation of the vertical displacement of the centre of mass between changes in sign of the 238 vertical acceleration using Equation 2 outlined above depends on knowledge of the fundamental 239 frequency of the acceleration signal generated from the loads recorded by the floor transducers. The 240 FFT of the raw load signal per horse indicated a peak amplitude below 2 Hz, and estimation of the 241 fundamental frequency of the signal by dividing half the number of reversals in the acceleration signal 242 by the total number of samples yielded fundamental frequencies of 1.94 Hz and 1.65 Hz for the two 243 horses, which roughly corresponds to the frequency of the peak amplitude in the FFT. The output 244 from that process would assume roughly even numbers of raw samples contributing to the averaged 245 positive and negative acceleration values, and although this is not likely per reversal, it is reasonable 246 to assume roughly even numbers across the entire trial. Given the trip started and ended in the exact 247 same place, and that the sum of all the actual calculated positive and negative displacements worked 248 out to -0.001 m and -0.002 m for the two horses, our findings suggest that this method of calculating 249 displacement is valid. The difference in calculated vertical displacement between the two horses is 250 accounted for by the difference in their fundamental frequencies. Theoretically, a stiffer horse will 251 have a higher vibrational frequency, whereas a horse that is better able to accommodate the vibrations 252 through its limbs will demonstrate reduced frequency and larger vertical displacement and mechanical 253 work as a result. The characterization of the differences may provide the means to quantify individual 254 mechanical responses to transport that have been reported previously [11,12]. 255 The method outlined here is proposed as a starting point for estimating the work done (ie. as a 256 proxy for energy expenditure) by the horse and other livestock during transport. In its simplest form, 257 Work = force x displacement. Working backwards from the average positive and negative 258 acceleration values generated by the pivot table, force (N) oscillates about the value for (mass of the 259 horse x gravity) and vertical displacement is calculated from the acceleration values per sign change. 260 Multiplying these average force values by the absolute vertical displacement values as the horse 261 bounces up and down in response to perturbations by the road surface yielded work values of 322.4 kJ 262 for Horse 1 and 443.2 kJ for Horse 2 over the course of the 58-minute trip. It is important to note that 263 this is a calculation for mechanical work, and there is not a 1:1 relationship with metabolic work, 264 although the two are linearly related during medium speeds of locomotion [14,15]. In real terms, there 265 is also mechanical work done by the horse making postural adjustments in the horizontal plane, and 266 these horizontal excursions were 227.1 m and 243.0 m respectively. The external horizontal forces 267 were not measured by our floor transducers, and so calculation of the horizontal work done is not 268 possible with this method. Work is done by the hip abductors and adductors to control movement in 269 the transverse direction, and also by the adductors of the forelimb. The flexors and extensors of the 270 proximal limb joints control motion in the cranio-caudal direction. Work done by these muscle groups 271 in controlling the transverse and sagittal motion of the trunk’s centre of mass would need to be 272 calculated from the horizontal forces against the floor, or by a musculoskeletal model with knowledge 273 of the moment arms and muscle forces. 274 It is not possible to calculate the mechanical work done against gravity in a static, standing animal 275 because there is no vertical displacement, aside from the very small changes in centre of mass position 276 caused by breathing. Calculating mechanical work from the vertical displacement of the centre of 277 mass requires a non-zero value for displacement, and so the negative displacements were made 278 absolute such that all the positive and negative changes contributed to a total displacement value that 279 was positive. However, the metabolic demand of positive work is greater than for negative work by at 280 least a factor of two [16,17] in humans performing gross cyclic motions like going up and down stairs 281 and cycle ergometry and the discrepancy increases with increasing speed. Whether this holds true for 282 small and transient alternating positive and negative vertical displacements caused by perturbations or 283 vibrations would be difficult to test using physiological methods. Burdett et al. [14] found good 284 correlation between increases in mechanical and metabolic energy cost at five gait speeds in humans 285 but did not investigate standing. Minetti et al. [15] determined that the rate of energy consumption of 286 standing horses was 1.94 ml O2 kg-1 min-1 and assumed that 1 ml of O2 is equivalent to 20.1 J for 287 conversion of metabolic into mechanical units during locomotion [18]. If that is the case, then the 288 mechanical cost of standing for Horse 1 in our study, over the course of 58 minutes would be 289 calculated at 1045 kJ, and the additional cost of 322.4 kJ accommodating the vibrations would be 290 additive. However, we simply added the absolute displacement values together without accounting for 291 the theoretically lower metabolic cost of the negative work. Also, the work done during these small 292 amplitude movements by series and parallel elastic elements in the muscles with little metabolic cost 293 would be impossible to quantify. 294 There are drawbacks to using a centre of mass approach for calculating work. The floor 295 transducers only measured the overall effect of all the sources of positive and negative muscle work 296 on the vertical displacement of the centre of mass as it moved in response to perturbations in the road 297 surface. It is likely that it would underestimate the metabolic effects of co-contraction by muscles in 298 near-isometric stabilisation of the limbs, and the simultaneous positive and negative work done by 299 muscles at different joints [19]. Work can also be calculated from changes in kinetic energy of the 300 centre of mass, but as we calculated vertical displacement from variable numbers of points in the raw 301 acceleration data, it would be difficult to calculate velocity with reasonable accuracy, and this is also 302 true of attempting to get velocity by integration of the acceleration curve. 303 Other factors influencing the energy cost of transport would be heart rate, and increased metabolic 304 rate caused by stress. This is a first effort to quantify the additional mechanical load caused by 305 adjustments the horse needs to make in response to perturbations caused by changes in direction, 306 speed and road surface. We will use it in future to evaluate the effects of the horse’s position in the 307 trailer (facing front, back, angled), along with other modalities like heart rate monitors and video. 308 309 Acknowledgements 310 311 This work was supported in part by operating funding from the Lewis Fitch fund and by a Major 312 Equipment grant from Massey University, Palmerston North, NZ. Brooke Gordon was supported by a 313 scholarship from the NZ Equine Trust. 314 The software developed in LabView for collection of the floor transducer and accelerometer data was 315 written by Dr. Jingpeng Wang. Mr. Mike Reilly and Mr. Andrew Smith of Massey University’s Large 316 Animal Teaching Unit were instrumental in the provision of the horses used in the study and assisted 317 with loading and unloading. 318 319 Footnotes 320 321 1. Paul-Douglas Floats, Angove Engineering Ltd., 17 Mangahao Road, Pahiatua 4910, New Zealand 322 2. Model ASB-250, PT Global Ltd, Auckland, New Zealand 323 3. National Instruments New Zealand Ltd., PO Box 147801, Ponsonby, Auckland 1021, New 324 Zealand 325 4. Model 356A32, PCB Piezoelectronics, ThermoFisher Scientific, 4 Talavera Rd, North Ryde NSW 326 2113, Australia 327 5. Microsoft Office 365 ProPlus, Microsoft Corporation. 328 6. 2019 Hilux SR5 2.8L 4WD Automatic, Toyota New Zealand. 329 330 Author contributions: C.B.R. and G.R.C designed the trailer and the experiment; C.B.R, G.R.C., 331 B.R.A, B.I.G. and B.E.M. collected the data. C.B.R., G.R.C. and L.T. analyzed and interpreted the 332 data. C.B.R. and G.R.C. wrote the article. All the authors reviewed the manuscript. 333 334 References 335 336 [1] Nielsen BL, Dybkjær L, Herskin MS. Road transport of farm animals: effects of journey duration on 337 animal welfare. Animal 2011;5:415–427. 338 [2] Statistics New Zealand. 2012 Agriculture Census Tables. 339 http://www.stats.govt.nz/browse_for_stats/industry_sectors/agriculture-horticulture-forestry/2012-340 agricultural-census-tables/livestock.aspx#livestock. Accessed 15Aug2019. 341 [3] Roy RC. Welfare of horses transported to slaughter in Canada and Iceland: assessment of welfare issues 342 and associated risk factors. Ph.D. dissertation. University of PEI, Charlottetown, Canada; 2014. 343 [4] Gonzales LA, Schwartzkopf-Genswein KS, Bryan M, Silasi R, Brown F. Relationships between transport 344 conditions and welfare outcomes during commercial long-haul transport of cattle in North America. J 345 Anim Sci 2012;90:3640-3651. 346 [5] Riley CB, Noble BR, Bridges JS, Hazel SJ, Thompson K. Horse injury during non-commercial transport: 347 Findings from researcher-assisted intercept surveys at southeastern Australian equestrian events. Animals 348 2016;6:65. DOI: 10.3390/ani6110065. 349 [6] Riley CB, Rogers CW, Padalino B. Effects of vehicle type, driver experience and transport management 350 during loading and in-transit on the welfare of road transport horses in New Zealand. In: Proceedings of 351 the New Zealand Society of Animal Production, 78, Lincoln NZ, July 2018, 92-95. 352 [7] Stull CL Responses of horses to trailer design, duration, and floor area during commercial transportation to 353 slaughter. J Anim Sci 1999;77:2925-2933. 354 ]8] Padalino B, Rogers CW, Guiver D, Thompson KR, Riley CB. A survey-based investigation of human 355 factors associated with transport related injuries in horses. Frontiers in Veterinary Science 2018;5. 356 http://doi.org/10.3389/fvets.2018.00294 357 [9] Preedy DF, Colborne GR. A method to determine mechanical energy conservation and efficiency in equine 358 gait: a preliminary study. Equine Vet J Suppl 2001;33:94-98. 359 [10] Lewis GS, Preedy DF, Colborne GR. Effect of variations in morphological data on calculated mechanical 360 energy cost in equine gait. Vet J 2002;164:71-73. 361 [11] Smith BL, Jones JH, Carlson GP, Pascoe JR. Body position and direction preferences in horses during road 362 transport. Equine Vet J 1994;26:374-377. 363 [12] Toscano MJ, Friend TH. A note on the effects of forward and rear facing orientations on movement of 364 horses during transport. Appl Anim Behav Sci 2001;73:281-287. 365 [13] Service d’Etudes Techniques des Routes et Autoroutes (Sétra). Assessment of vibrational behaviour of 366 footbridges under pedestrian loading. Ministry of Transport and Infrastructure, France, 2006. 367 http//:www.setra.equipement.gouv.fr 368 [14] Burdett RG, Skrinar GS, Simon SR. Comparison of mechanical work and metabolic energy consumption 369 during normal gait. J Orthop Res 1983;1:63-72. 370 [15] Minetti AE, Ardigo LP, Reinach E, Saibene F. The relationship between mechanical work and energy 371 expenditure of locomotion in horses. J Exp Biol 1999;202:2329-2338. 372 [16] Abbott BC, Bigland B, Richie JM. The physiological cost of negative work. J Physiol 1952;117:380-390. 373 [17] Beaven CM, Willis SJ, Cook CJ, Holmberg H-C. Physiological comparison of concentric and eccentric 374 arm cycling in males and females. PlosOne 2014;9: e112079. DOI: 10.1371/journal.pone.0112079 375 [18] Blaxter K. Muscular work. Energy metabolism in animals and man. Cambridge: Cambridge University 376 Press; 1989, p. 147-171. 377 [19] Winter DA, Mechanical work, energy and power. In: Biomechanics and Motor Control of Human 378 Movement. 3rd ed., Hoboken, NJ: Wiley & Sons, 2005, p. 118-155. 379 380 381 382 383 (a) 384 385 (b) 386 387 (c) 388 389 390 Fig. 1. Fast Fourier Transform (FFT) on (a) the load data collected from the steel bench transport 391 trial, on (b) the vertical acceleration data collected from the bulkhead-mounted accelerometer and on 392 (c) the load data collected from a horse transport trial. 393 394 0 1 2 3 4 5 6 7 0 5 10 15 20 FF T M ag n it u d e FFT Frequency (Hz) 0 0.05 0.1 0.15 0.2 0.25 0.3 0 5 10 15 20 FF T M ag n it u d e FFT Frequency (Hz) 0 2 4 6 8 10 12 14 0 5 10 15 20 FF T M ag n it u d e FFT Frequency (Hz) 395 396 Fig. 2. Vertical displacement of the horse’s centre of mass, calculated from the acceleration data. 397 Load data were collected at 100Hz, filtered with a low-pass Butterworth filter at a cut-off frequency 398 of 5Hz and then the number of reversals in sign of the acceleration signal determined with a pivot 399 table. In this case, the number of reversals in sign was 13,478 from 346,916 samples at 100Hz. 400 401 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 1 1001 2001 3001 4001 5001 6001 7001 8001 9001 10001 11001 12001 13001 D is p la ce m en t (m ) Time (Reversals) 402 403 Fig. 3. Horizontal excursion of the horse’s centre of pressure during transport. Raw data collected at 404 100Hz were filtered using a low-pass Butterworth filter at a cut-off frequency of 5Hz and then every 405 100 frames were averaged to yield one point per second. 406 -1600 -1400 -1200 -1000 -800 -600 -400 -200 0 0 200 400 600 800 Fo re -a ft p o si ti o n ( m m ) Transverse position (mm)