SEMESTER PROJECT Design and implementation of a force/torque sensor for a quadruped robot Supervisers : Alexander Spröwitz Rico Möckel Prof. : Auke Jan Ijspeert Nicolas Sommer, master MT 20/06/2011
PRESENTATION OUTLINE Introduction Cheetah/Oncilla and CPGs : sensory feedback Electronics Sensor design Simulation of sensor’s deformation Results Conclusion/questions 2
INTRODUCTION Two-parts semester project : • Analyze and possibly correct previous project on the Roombots • Develop a multi-axis sensor for the Cheetah-robot Roombots Old Cheetah prototype 3
CHEETAH AND CPG S: INTEGRATION OF SENSORY FEEDBACK • Cheetah + CPG : Gait up to 1m/s [1] • It is possible to use sensory information in a CPG so that the oscillator is better coupled with the mechanical system [2] • 6 axis sensor gives the GRF correlate with phase of the gait • Enables to detect collision [1] A. Tuleu, A. Sproewitz, M. Ajallooeian, P. Loepelmann, and A. J. Ijspeert. Exploiting Compliance with a Cat-sized Quadruped Robot for Trot Gait Locomotion. *Biomechatronics, TU-Ilmenau, Germany. Biorobotics Laboratory, EPFL, Lausanne, Switzerland. [2] L. Righetti and A. J. Ijspeert. Pattern generators with sensory feedback for the control of quadruped locomotion. Proceedings of the 2008 IEEE International Conference on Robotics and Automation (ICRA 2008), Pasadena, May 19-23, 2008. 4
DATA ACQUISITION ELECTRONICS Wheatstone bridges Analog voltage about |DC|<200mV dsPIC COMPUTER USB/serial (Matlab) SPI ( UART ) (microcontroller) Analog-Digital COM PORT Formatted Converter Digital Values, home- 24-bits values made protocol Debugging the SPI communication 3 software interfaces : Clock • PC : Chip select • Matlab Data out (SDO) Data in (SDI) • dsPic : • UART PC • SPI ADC 5
SENSOR DESIGN FOR QUADRUPED ROBOT Hypothesis : 1. Three forces between foot and floor (GRF) 2. No moments transmission. 3. Position of contact (below: A) varies and position of the foot unknown 𝐺 0 𝑦 z 𝐺 0 𝒰 𝑔𝑚𝑝𝑝𝑠→𝑔𝑝𝑝𝑢 = 𝑧 y y AB . 𝐺 0 𝑩 x 𝑨 𝐺 𝑁 𝑦,𝐶 = 𝐺 𝑧 ∗ 𝑨 𝐵𝐶 − 𝐺 𝑨 ∗ 𝑧 𝐵𝐶 B 𝑦 𝐺 𝑁 𝑧,𝐶 = −𝐺 𝑦 ∗ 𝑨 𝐵𝐶 + 𝐺 𝑨 ∗ 𝑦 𝐵𝐶 = z AB 𝑧 𝐺 𝑁 𝑨,𝐶 = 𝐺 𝑦 ∗ 𝑧 𝐵𝐶 − 𝐺 𝑧 ∗ 𝑦 𝐵𝐶 𝑨 Foot A 𝑪 Useful to define dimensions and Third segment and foot schematic to compute forces in A from B A : contact between foot and floor B : position of the sensor (third leg segment) 6
SENSOR DESIGN FOR QUADRUPED ROBOT (2) • Choice of a 6-axis sensor • No restriction on the location of the contact • Simplicity of the design • Design inspired from robot’s finger 6 -axis sensor [1] • Small, already tested and very precise • Same range of forces • One Wheatstone bridge sensitive to only one force/moment component by design [1] G-S Kim 2004 Development of a small 6-axis force / moment sensor for robot’s fingers. 7
SENSOR DESIGN FOR QUADRUPED ROBOT (3) • Sensitivity and resistance considerations to determine the dimensions • Specifications on minimum sensitivity for each component • Maximum forces admissible Reduce stress concentrations • Third part for a better access to glue the gauges. • Rounded edges to avoid stress concentrations • Attachement points at the extremities Prototype parts 8
DEFORMATIONS FROM FORCES Gauges Each picture shows the deformation along the axis of the drawn gauges 9
DEFORMATIONS FROM MOMENTS Values sampling done in the middle of each gauge but tested on larger surfaces and little difference. 10
∆L 𝑊 𝑝𝑣𝑢 = 𝑊 𝑗𝑜 ∗ 𝐿 ∗ 4 SIMULATION RESULTS Full bridge configuration - gauge factor K = 150 - Bridge V in =5V Bridges-to-force voltage Bridges-to-torque voltage 20 600 15 400 Output Voltage (mV) Output Voltage (mV) 10 200 1,93 mV 5 0 0 B1 B2 B3 B4 B5 B6 B1 B2 B3 B4 B5 B6 -200 -5 -400 -10 -15 -600 Fx=1N Fy=1N Fz=1N Mx = 1N.m My = 1N.m Mz = 1N.m • Each bridge (from B1 to B6) measures one force/moment component only • Good output ratios • Low sensitivity on Fz 𝟐 𝟓 𝑶𝒇𝒙𝒖𝒑𝒐 (25g) with 0,5mV reading precision • Still a Fz resolution of about 11
∆L 𝑊 𝑝𝑣𝑢 = 𝑊 𝑗𝑜 ∗ 𝐿 ∗ 4 SIMULATION RESULTS Full bridge configuration - gauge factor K = 150 - Bridge V in =5V Bridges-to-force voltage Bridges-to-torque voltage 20 600 15 400 Output Voltage (mV) Output Voltage (mV) 10 200 1,93 mV 5 0 0 B1 B2 B3 B4 B5 B6 B1 B2 B3 B4 B5 B6 -200 -5 -400 -10 -15 -600 Fx=1N Fy=1N Fz=1N Mx = 1N.m My = 1N.m Mz = 1N.m • Each bridge (from B1 to B6) measures one force/moment component only • Good output ratios • Low sensitivity on Fz 𝟐 𝟓 𝑶𝒇𝒙𝒖𝒑𝒐 (25g) with 0,5mV reading precision • Still a Fz resolution of about 12
EXPERIMENTAL SETUP • Half-bridges Time constraint - Only drawback = sensitivity • Only Fx and Mz • Acquisition with Labview
EXPERIMENTAL TESTS Bridge n°1 Output voltage (mV) Bridge n°6 Output voltage (mV) 50 200 180 mV 44 mV 172 mV 45 41 mV 180 Output voltage (mV) Output voltage (mV) 40 160 35 140 30 120 25 100 20 80 15 60 10 40 5 20 0 0 Fx=10N Mz=1N.m Simulation Experiment Simulation Experiment Linearity not tested yet, possible causes : • Good results ~180mV • Gauges non-linearity, specified values : ∆𝑀 𝑀 < 6E-4 ) • Fx error < 7% • Better than ±0.25% to 600 µm/m ( ∆𝑀 • 𝑀 < 1.5E-3 ) Better than ±1.5% to 1500 µm/m ( • Mz error < 5% • Elastic limit reached much higher deformation 14
CONCLUSION • Design of a 6-axis force/moment sensor that fits in the robot’s leg’s third segment and is lightweight (~15g) • Separation of the forces/moments components on each bridge and good output ratios • First experimental results match simulations well (5 and 7%) • Improvements : Complete electronics, run resistance tests, increase FEA part precision , improve test setup • Next work : Make use of the data QUESTIONS 15
Recommend
More recommend
