Robotic Wheel Drive means controlling wheels used to propel robotic devices such as warehouse robotics, rovers, agricultural robots, robotic lawn mowers and a wide array of mobile devices. Some such devices have two primary drive wheels but three, four and even higher numbers of drive wheels are possible. The unique motion control application that ties these machines together is the challenge of slip control and navigation over uneven and physically diverse terrain.
The motion control challenge involved with wheel-driven mobile devices differs in one very important respect from virtually all other motion control problems: wheels are not solidly attached to the surfaces they contact and rely on friction for creating motion. This means controlling the motor's angle position, velocity, or torque output doesn't allow the controller to perfectly predict the location and trajectory of the mobile device.
The primary reason for this is wheel slip, which means the surface of the wheel loses contact with the surface it is travelling on. A secondary reason is that the surface itself may be bumpy and uneven. The net result is similar - even if the wheels precisely rotate a distance that should result in (for example) one meter of robot movement, that actual distance travelled may be less or sometimes more. And even if the robot wheels are commanded to turn the robot to a precision direction it may end up pointing in a different direction.
The most common motion technique to combat this is to drive the wheels with a velocity loop controlled by an outer loop position controller. A velocity loop is critical for keeping wheels synchronized and moving at the desired speed. Compared to a position loop used for the same task, a velocity loop won't suffer from the problem of multiple wheels 'fighting each other' as a result of small surface skips and bumps. This is because velocity loops don't 'remember' and try to restore position.
The outer loop controller measures the robot's orientation and location and generates, for each wheel, a commanded angular velocity (again depending on the wheel configuration) to change the robot's direction or speed so that the system motion matches as closely as possible the planned robot travel path. GPS is a common method for determining the robot's position, but other devices such as magnetic guide tracks, cameras or lidar may be used. Gyroscopes are sometimes used to supplement these sources if highly accurate orientation information is needed.
A common additional approach to detecting and correcting for wheel slippage is use of a motion observer. Kahlman filters and Luenberger observers are examples of algorithms that may be used to input information from multiple controlled wheels to infer when a specific wheel has slipped. If the amount of slip is known it can be used to adjust a 'dead reckoning' model of robot movement to supplement GPS or other location data.
Torque limiting is an another important motion technique that is used to reduce the probability of slippage. If the wheel torque slippage threshold is known (this is the commanded wheel torque value beyond which it may lose proper contact with the surface and slip) this threshold value, reduced by 20 or 30%, can be programmed in to the controller to reduce slippage.
Finally being mobile, wheel drive applications require miniaturization of the drive electronics to reduce weight, and require modern high efficiency drives to extend battery life. Luckily today's motion controllers are getting smaller and more powerful every year, so this requirement is becoming easier and easier to satisfy.
The diagram below shows a generalized outer loop controller which uses the qa (actual robot position to generate the linear and angular velocities the robot has to follow to reach the trajectory). This robot motion information is transformed to velocity commands for each driven wheel and output to a velocity loop. The velocity loop, one for each driven wheel, inputs via an encoder the wheel position from which it derives the velocity and then calculates a velocity error which is passed through a PI (proportional, integral) filter and output as a torque command to the amplifier.
Since 1994 Performance Motion Devices products have been used in a range of mobile applications including welding inspection robots, agricultural robots, bomb disposal robots, construction robots, rovers and more. PMD’s ultra-compact and powerful N-Series ION Drive is especially well suited for robotic wheel control, as are the MC58113 IC and Juno Family of ICs.
PERFORMANCE MOTION DEVICES, INC.
80 Central Street | Boxborough, MA 01719
P: 978.266.1210 | F: 978.266.1211 | info@pmdcorp.com