A Self-contained 3D Hopping Robot Kale Harbick Department of Computer Science, USC
2 Objectives Self-contained –Onboard computing –Onboard power –Capable of dynamically stable motion without external support Additional behaviors –Gait transitions (inclined planes) –Sit down, stand up, lean against a wall
3 Static vs. Dynamic Stability Statically stable –Car –Quadruped walking Dynamically stable –Biped walking –Running, hopping
4 Previous Work Marc Raibert – MIT Leg Lab –Planar and 3D hoppers –Planar and 3D bipeds –Quadruped (trot, bound, pace) Martin Buehler – McGill University –Monopod I and II –Scout I and II (bounding quadrupeds)
5 Previous Hoppers
6 Three-part Control System Raibert controller Hopping height –Thrust for specified duration during stance –Exhaust to specified pressure during flight Forward velocity Body attitude
7 Forward Velocity
8 Body Attitude
9 My Design
10 Mechanical System Leg thrust –Pneumatic cylinder (400N) Leg swing –2 Pneumatic cylinders (250N) Power –2 onboard CO 2 tanks (5 min) 9 kg total mass
11 Processing and Communication Processing –486DX133 PC/104 –Solid state disk –A/D Converter –Quadrature Decoder Communication –Radio modem –Infrared
12 Sensors Foot contact 2 encoders for hip angle Leg length 3-axis accelerometer Roll and pitch gyros Compass Pressure sensors
13 Performance Predictions 15 cm foot clearance 1.1 m/s maximum velocity 135 W (mechanical) Specific resistance = 1.4
14 Performance Comparison MassPowerSpecific Resistance MIT 3D Hopper 17 kg250 W1.5 Monopod I 15 kg125 W0.7 Monopod II 18 kg48 W0.22 MIT Quadruped 25 kg2500 W5.0 USC 3D Hopper (predicted) 9 kg135 W1.4
15 Specific Resistance P = mechanical power m = system mass g = acceleration due to gravity v = forward velocity
16 Specific Resistance
17 Additional Behaviors Sitting down, standing up, leaning Joystick control –Turn by changing direction of motion –Yaw is not controllable Inclined terrain –Up vs. Down