How do mobile Robots move? (Thu Sep 12, lect 5) | previous | next | slides |

Wheels, legs and other ways for robots to move around 3D space

Logistics

Steering and navigation

  • Focusing for now on autonomous, ground based mobile robots.
  • Software needs to be able to command the robot
  • Given wheels with motors, or tracks, or casters, or legs
  • Electricity controlled through software makes them move
  • So consider this…

Concept: Degrees of Freedom

DOF: How many does each one have?

  • Train?
  • Baseball?
  • Drone?
  • Car?
  • Tank?

DOF: Considerations

  • How many axes can move
  • Within what is physically possible
  • Degrees of freedom:
    1. Vehicles
    2. But also grippers, arms, doors, anything that can move
  • Subtleties of you look for them…
    1. Consider which coordinate system
    2. Fixed attachments
    3. Single instantaneous motion

Locomotion: Holonomic vs. Non-Holonomic

  • Possible vs. Controllable degrees of freedom
  • If they are the same, then Holonomic
  • If they are different, then non-Holonomic
  • Car
    1. Will never leave the ground
    2. So it’s actual DoF is 3 (move forward and sideways, and yaw)
    3. But its controllable DoF is 2 (wheel rotation and steering)
  • Differential Drive
    1. Actual DoF is 3 also
    2. Controllable DoF is 1 (wheel rotation)

Different Configurations for wheeled

  • “Differential drive”
    1. Steering is achieved by driving wheels at different speeds
    2. Geometry calculations to determine what delta rpm leads to what kind of turn
    3. Usually three wheels, two with motors, one “caster” (or two)
    4. Different kinds of caster make a difference (I found this out)
  • Tri-cycle steering
    1. Two fixed Wheels
    2. A third wheel with a steerable mount
  • Ackermann Steering
    1. Where the steering is achieved by actual rotation of the wheel mount
    2. Note subtle difference from parallel steering
  • Four wheel drive
  • Articulated Vehicle
    1. Where the vehicle itself has a pivot point
    2. Notice that to steer the vehicle an algorithm needs to convert the desired rotation to proportional wheel spinning

Firmware

  • Software and hardware architecture
  • cmd_vel: needs to convert from high level units to motor commands
  • odom: needs to convert from sensors (motor decoder and others) into coordinates
  • Big deal is being aware of known errors from the manufacturer
  • We wont go into the detailed geometry and math but it is very well studied and understood

What is needed (data, sensors, algorithms)

  • Discuss in groups. Pick one of these problems and your approach to them.

    1. Drive in a straight road: how do you detect the road? How do you stay on your side? What sensors are needed?
    2. Turn a corner: How do you identify a corner? Left vs Right Corner? What about a hallway? What about a dead end?
    3. Make a 90 degree turn: Without any landmarks, in an open field? Is left or right different?
    4. Stay in a lane: How do you identify the lane? What about road conditions? What about night vs. day?
    5. Steer from Shapiro to Usdan: How do you designate where Usdan is? What other information do you require?
    6. Move at x Meters/second: Make it as precise as possible; even if the road is slippery or there’s a steep hill

Introducing Kinematics

  • In general: given a pose, and a series of commands, calculate the new pose (forward kinematics)
  • And, given a pose and a desired new pose, what are the series of commands that will get you there (inverse kinematics)
  • Note that the concept and terminology apply equally to vehicles and to other structures
  • Requirement:
    1. An initial pose
    2. A final pose
    3. A series of commands
    4. A way to calculate the relationship
  • For a differential drive, the math is complex but well understood
  • Note the relationship to path planning

Other scenarios

  • Legged Vehicles
  • Tracked vehicles (tanks)
  • Water surface vehicles
  • Under water vehicles
  • Aerial vehicles

Terrain

  • Indoor vs. outdoor
    • Traction and slippage
    • Weather effects - wet, dry, light, dark
    • Visibility of Sky
  • Inclines and declines and surface
    • Needs more force
    • Also affects levelness
    • Direction that cameras and sensors are pointed

The role of sensors in controlling motion

  • Self-sensors (like odometry)
  • External sensors (like LIDAR)

Thank you. Questions?  (random Image from picsum.photos)