Mobile robots. Structure of this lecture. Section A: Introduction to AGV: mobile robots. Section B: Design of the ShAPE mobile robot

Transcription

1 slide n. 1 Mobile robots Development of the ShAPE mobile robot Ing. A.Tasora Dipartimento di Ingegneria Industriale Università di Parma, Italy tasora@ied.unipr.it slide n. 2 Structure of this lecture Section A: Introduction to AGV: mobile robots Section B: Design of the ShAPE mobile robot 1

2 slide n. 3 Section A: Introduction to AGV: mobile robots slide n. 4 Autonomous guided vehicles Mobile robots: used for surveillance logistics entertainment, etc. Solutions are different in terms of method of locomotion (wheels, legs, tracks, etc.) payload & speed (performance) navigation system etc 2

3 slide n. 5 Some ready-to-use AGV Esatroll Paquito Max speed 1.3 m/s with laser scanner and bumpers Proxaut MT10 Max speed 1.3 m/s Max payload 1000 kg LGV navigation, with laser and gyroscope slide n. 6 Some ready-to-use AGV Skilled MT10 Max speed 1.5 m/s Max payload 2500 kg LGV navigation, with laser Repeatability: 10 mm 3

4 slide n. 7 Locomotion systems Propeller / jet / rocket.. (UAV, unmanned aerial vehicles) 6-DOF navigation GPS + gyroscopes + magnetic gyrocompass + vision awareness + laser altimeter + accelerometers (and Kalman filter ) Legs Difficult to control Useful for uneven pavements Not useful for industrial environments Tracks / snakes / etc. Mostly for research not in industry slide n. 8 Locomotion systems Three Interroll omnidirectional wheels No need to turn wheels: direct transmission with 3 motors All types of 3-DOF manouvers on 2D plane Not suited for high speeds Not suited for high loads Possible improvements ( Mecanum wheels) 4

5 slide n. 9 Locomotion systems 3 or 4 fully steerable wheels All types of 3-DOF manouvers on 2D plane Good performances but Complex design (more motors than DOFs) slide n. 10 Locomotion systems Two parallel wheels and one steerable wheel Simplified design Only two motors Good speed & payload (ex. industrial environments) Not all 3-DOF motions in 2D are possible! (non-holonomic constraints) Two different approaches: Differential wheels Motorized steering wheel m1 m2 m1 m2 5

6 slide n. 11 Locomotion systems Motorized steering wheel m1 m2 Advantages: Front wheel never gets stuck Disadvantages Two sizes for the motors One of the two motors works much more than the other The mechanism for steering requires vertical space slide n. 12 Locomotion systems Differential wheels m1 m2 Advantages: Same size for motors, reducers and controllers Both motors are used for accelerating lightweight design Very simple to build Small footprint Disadvantages The front wheel has passive steering, it can get stuck.. 6

7 slide n. 13 Navigation systems & sensors How to get the absolute position (x, y,θ ) of the robot? Odometric data (recordings of wheel rotation) is not enough! It accumulates errors it must be integrated with other more absolute information.. Y I Y R G X R θ Absolute position must be updated in real-time, as fast as possible X I No need for extreme precision (10 mm repeatability is good) Solution? Different systems are used. slide n. 14 Navigation systems & sensors Robot on railways / on guides Easy solution, but not flexible Requires expensive modifications to the building floor/roof Wires in the floor & inductive sensor Easy solution, not 100% flexible Requires expensive modifications to the building floor Optical lanes painted on the floor Easy solution, not 100% flexible Cheap modifications to the building floor, but painted lines on the ground can be covered by dirt 7

8 slide n. 15 Navigation systems & sensors Gyroscopes Only rotation information Mechanical / Laser Sagnac effect / Piezo (MEMS) Only piezo gyros are cheap, but easily accumulate drifting.. Magnetic gyrocompasses Only rotation information Extremely cheap (two IC fluxometers) Measure the magnetic field of Earth absolute, but low precision Affected by disturbs slide n. 16 Navigation systems & sensors Satellite GPS Only x,y position Not precise enough (but cheap) Requires open air MEMS gyroscopes + MEMS accelerometers ( + gyrocompass + ) 3 DOF rotation without drifting Useful for attitude of UAV, drones, etc Redundant sensors: exploit Kalman filters Adding GPS for translation too: full 6 DOF 8

9 slide n. 17 Navigation systems & sensors Example: a quadcopter drone with autopilot (Ilmenau University, DE) slide n. 18 Navigation systems & sensors Laser navigation (LGV) Both x,y position and rotation Very used for industrial AGV Rotates a laser and sees when it hits some fixed reflective markers in the building Problems with occluded markers / bad illumination Not that cheap 9

10 slide n. 19 Navigation systems & sensors Feedback with artificial vision 1) One or more camera on the roof see the AGV 2) Image analysis software can extract features from camera views 3) Position of AGV is obtained in view field, then trasformed to abs.space No need to put the computer on the robot Often used for small robots (soccer robot games, etc) Robots must have recognizable symbols on their top (problems with bad illumination, etc.) Artificial vision awareness (SLAM approach) The camera is mounted on the robot the robot looks at the environment which it navigates, while an AI software with artificial vision can understand the position respect to known objects (walls, windows). Very complex sw, low robustness not ready for industrial applications. slide n. 20 Section B: Design of the ShAPE mobile robot 10

11 slide n. 21 Operating environment The robot must carry small boxes filled with plastic materials Small footprint is required (max 1m length) No need to buy large commercial AGV We developed a custom AGV, with simple navigation method based on feedback from fixed videocameras and image analysis slide n. 22 Operating environment The carthesian robot which assists the AGV and the storage system 11

12 slide n. 23 Operating environment The storage system: how the load/unload buffer works slide n. 24 Locomotion system Differential wheels m1 m2 We choose the differential system because, among other advantages, allowed us to keep the vertical size of the load plane under the strict requirement (150 mm) 12

13 slide n. 25 Overall sizing slide n. 26 Choosing motors and transmissions Requirements: Speed: 1 m/s Ramps: 8% Accelerations: as from various benchmark for typical duty cycles.. Results: Reducers ratio: 1/20 Wheel diameter: 120 mm Brushless motors LENZE Fluxxtorque 931E (0.8 Nm nominal torque) 13

14 slide n. 27 Choosing motors and transmissions Lenze brushless motors (24V) slide n. 28 Choosing motors and transmissions The worm reducer Low precision (some backlash) and low efficiency but..fits into budget constraints..and takes small room in the robot frame 14

15 slide n. 29 Mechanical design The aluminum truss slide n. 30 Mechanical design The box for drive controllers, electronic devices and accumulators 15

16 slide n. 31 Mechanical design Details slide n. 32 Mechanical design The wheel: it must touch the ground in a point (i.e. the smallest possible area) Bearings must be resistant (1000N of radial force) 16

17 slide n. 33 Mechanical design The pivoting wheel must be as stiff as possible, with toroidal surface, so that it does not create unwanted frictional effects during changes of direction. We tried different types of materials. Cast polyurethane is worse than hard polyammide. Cylindrical tire is worse than beveled or toroidal surface. slide n. 34 Electrical design The 24V power circuit 17

18 slide n. 35 Electrical design The accumulators: 4 x 27Ah standard lead batteries Predicted continuous operating time without need to recharge: 2h. slide n. 36 Control The AGV is controlled by a remote computer using Wi-Fi ethernet The remote computer is fixed to ground (it does not waste electric power) while on the AGV there are only simple controllers for the simplest tasks The remote computer is also responsible of complex image analysis from the fixed videocamera 18

19 slide n. 37 Control Hi-Res Videocamera firewire Remote computer Router wireless Bridge wireless Converter Ethernet CAN MCU realtime controller Drives of the two motors CAN bus Ethernet Wireless IEEE g Ethernet slide n. 38 Control Bridge wireless Converter ethernet/can Note!!! This CAN-over-WiFi scheme is enough for the prototype, but NOT for hard-real-time environments (an embedded controller should take care of RT) 19

20 slide n. 39 Control The two drives for the control of the brushless motors slide n. 40 Software The software updates the state of the robot each 20ms Acceleration / speed / rotation ramps for the two wheels are calculated on-the-fly, so the speed setpoint is continuously passed to the two controllers with CAN telegrams: 20

21 slide n. 41 Software The user interface Allows: - jogging - storing a position list - programming slide n. 42 Software Example of program running through a position list 21

22 slide n. 43 Repeatability Test: good results even with open-loop feed-forward only slide n. 44 Examples 22

23 slide n. 45 Conclusions The ShAPE mobile robot is a custom AGV with good performance and low cost Global positioning comes from artificial vision CPU-intensive operations are performed on a computer that is fixed to ground Information passed to the AGV using Wi-Fi devices. 23