Robotic Vehicle Design Sensors, measurements and interfacing Jim Keller July 19, 2005
Sensor Design Types Topology in system Specifications/Considerations for Selection Placement Estimators Summary
Sensor Types (position/motion sensing as an example) Position Relative vs. absolute GPS provides globally referenced position Rangefinders provide local position information with respect to the environment Characteristics (where do the sensors work well and where do they not, what important features must be taken into account GPS works well in open spaces; may not work at all in most urban areas Rangefinders may work well unitl environment becomes cluttered or uneven Velocity Groundspeed Air or water speed for aerial or marine robots Typically a local measurement relative to the robot itself Acceleration Almost always an absolute measure unless calibrated Orientation (pitch/roll/compass heading) Rotational rate Integrated packages are expensive: Very low end ~ $1500 Precise ones > $100K
Topology Sensors provide feedback about the environment Delay imposed by sensor or sensor processing must be considered if response time is important Input Device Reference Commands (if tele-operated) System Processor Command Processing Control Actuators Unaugmented Vehicle Action/ Response Feedback Sensors
Specifications/Design Considerations Range Sensors have a limit on upper and lower bounds of the states they measure (i.e. temperature: -40 o F to 212 o F ) Precision/Accuracy/Resolution/Tolerances Precision is how repeatable are measurements when sensing the same state (not always the same as accuracy due to drift etc.) Accuracy is how close is the sensed value to the actual value of the sensed state Resolution is how finely can the sensor distinguish changes in state (i.e. temperature: output may only be available in 1 o F increments) Tolerances are bounds identified with respect to the above rating a sensor s performance (i.e. temperature: Accuracy +2 o F) Type Analog/Digital Analog sensors may have a digital interface but still exhibit idiosyncrasies of analog equipment (i.e. airspeed sensor may still be subject to temperature drift) Direct digital sensors do not drift (i.e. digital encoders for rotation or translation) Noise Noise cannot be distinguished from real data so signal to noise ratio for sensor is important Direct measure or estimation Estimation may be best but noise can limit utility Know the features of all sensors you use!
Dual/Frequency Split Sensors Technology may permit a single sensor meet requirements Multiple sensors can be integrated in a variety of filter configurations to provide a composite filter Example: Airspeed filter: Accelerometer estimates may be the best to know how quickly speed is changing but not the absolute speed Pressure sensing may be the best way to know the absolute speed but may be slow to respond to changes or subject to noise Solution: use pressure sensing to measure low frequency component of airspeed; acceleration to measure the high frequency component combine signals into a complementary filter so the robot uses one sensor
Sampled Data System Contraints (input frequency cannot exceed ½ sample frequency) 1 Aliasing occurs if data with a waveform higher than ½ the sample frequency is sampled High frequencies are aliased to lower frequencies Prefiltering is required to eliminate aliasing if input cannot be guaranteed to adhere to max frequency constraint 10 Hz signal 60 Hz signal 0.8 0.6 0.4 0.2 0-0.2-0.4-0.6-0.8-1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 Time - seconds Data sampled at 50 Hz (T = 0.02sec)
Static vs. Dynamic Compass Magnetic field lines are a 3D vector (North, East, Down) Handheld compass is meant to be held level by a stationary user Robotic sensors typically use 3-axis magnetometer to sense field lines in 3D Accelerometers can be used to correct for static orientation of down component Orientation sensor (not valid for moving platforms) Gyros needed to correct for motion Hence term gyro-compass You need to have the proper type of sensor for your own application
North Angle Parameter Definitions Present Position Course Last Waypoint Cross Track Error Track Angle Heading Bearing East Groundspeed Track Angle Error Desired Track Along Track Error Next Waypoint Notes (GPS can provide Track relative to True North) For ground vehicles: Heading = Track (except when skidding) Angular references for air vehicles come in many flavors (all are typically measured from North): Heading - in which direction am I pointed Track - in which direction am I moving Bearing - which direction should I move to get from where I am to where I want to go Course which direction should I plan to go to get from one point of interest on my route to the next point of interest
A magnetic sensor is biased by declination angle based on its location use online resources to determine (models available for download) Philadelphia declination is: -12 22' (W) Magnetic Heading Sensor True North = sensor reading - declination
Co-location Placement Always best to place sensor closest to location of importance Flexible structures typically require understanding of mode shapes Environment Heat Electromagnetic interference Cameras should not be next to motors etc
Estimators It may not be possible or even desirable to directly measure Position estimation for submarines A variety of state variables may be used in conjunction with a math model of the robot to estimate position based on other available measurements Estimators must use feedback to continuously update themselves Probabilistic model of system very useful to credibly integrate diverse signals into a single sensor Kalman filter developed in 1960 s continue to evolve today to integrate sensed data into a best estimate Time tagging of data critical when diverse sources of data are combined Estimation for stabilization should be used as a second resort Estimation introduces some delay or lag and may destabilize system to some extent
Summary Select sensors required for a robot to accomplish its task(s) Use estimation as required May not be required for some applications may be critical to others Understand the specifications and performance of a sensor before committing to use it in a robot Prototype robot with variations in preliminary design to make best sensor selection