Understanding Smart Sensors Sensor Fusion - Using the Newest Techniques for Advanced Sensors October 12, 2012 Randy Frank Agenda • Two Fusion Paths • Sensor Fusion Models • Applications – Fusing Radar & Camera Data – Motion Fusion • Sensor Fusion Products • Sensor Fusion Tools/ Dev Kits Sensor Fusion Decision making capability Sensor (data) fusion signal processing neural networks fuzzy logic multiplexing multiple types of sensors in same package multiple measurements of same type of sensor, i.e. 3axis accelerometer single parameter sensor Sensor output Direct sensor fusion Dasarathy Model Courtesy Belur Dasarathy Sensor Fusion Confusion Fusing Radar and Camera Inputs Sensor Fusion in Vehicles Source: Delphi Sensor Fusion for Virtual Sensing • NIRA Dynamics Autonomous Driving DARPA Challenge • Tartan Racing, Carnegie Mellon University Sensor Fusion for Motion Source: STMicroelectronics Sensor Fusion for Motion Pressure Trim hi/low/band pass filtering Pressure Shake detection shake event 3-Axis Acc FoR mapping Trim hi/low/band pass filtering 3-Axis Gyro FoR mapping Trim hi/low/band pass filtering ω x,y,z 3-Axis Mag FoR mapping Trim & Hard/Soft compensation hi/low/band pass filtering B x,y,z Raw data Acc x,y,z calibration parameters Calculate hard/soft iron parameters FoR = Frame of Reference Rotation matrix Kalman Filter or similar function Quaternion Geometric computations Tilt-compensated mag heading Inclination (φ, Θ, Ψ) Sensor Fusion MANY styles of sensor fusion are possible. Source: Freescale Medical Example Source: Freescale Motion Capture Systems • MVN inertial motion capture suit from Xsens • System’s 17 inertial trackers and the sensor fusion software allow motion capture without requiring cameras • Used in animation as well as medical and sports applications Sensor Fusion Products • Many sources for motion control – Analog Devices – Bosch – Freescale – InvenSense – Kionix – MEMSIC – STMicroelectronics 10-DoF MEMS IMU • Analog Devices ADIS16480 integrates: – tri-axis gyroscope – tri-axis accelerometer – tri-axis magnetometer – pressure sensor – ADSP-BF512 Blackfin® processor • Incorporates an extended Kalman filter (EKF) to fuse sensor inputs 10-DoF MEMS IMU • EKF takes multiple measurements over time, and merging them with a predictive state estimator. • Intensive code development, testing and external processing required by other MEMS IMUs. • Targets: military and commercial aircraft navigation, unmanned vehicles, movable platform positioning, and industrial robotics. • Evaluation board Motion Tracking Device InvenSense MPU-6500 • Turnkey 6-axis MotionTracking – MEMS gyroscope and accelerometer – Onboard Digital Motion Processor™ (DMP) – MotionFusion algorithm – 3x3x0.9mm QFN package Motion Tracking Device InvenSense MPU-6500 • Applications such as pedestrian navigation, context-aware advertising, and other locationbased services • Wearable sensor applications such as remote health monitoring, sports and fitness tracking, Architecting Fusion Source: Movea Design Options • Software suite operates either as a hardware solution on an embedded MCU or as a software solution on an AP Software Solution Sensor Platforms’ FreeMotion™ Library • Supported microprocessors include: • 32-bit embedded processors (ARM’s CortexM, Atmel’s AVR and Freescale’s ColdFire families used as sensor hubs) • 64-bit application processors (Intel’s Atom, nVidia’s Tegra, Qualcomm’s Snapdragon, and TI’s OMAP processors used in smartphones and tablets) Sensor Fusion Tools 12-Axis Sensor Platform • Complete hardware and software solution Other Development Kits • Sensor Platforms FreeMotion™ Library and Software Development Kit • SmartFusion Development – Actel • MTi Development Kit – Xsens • ADIS16480/PCBZ breakout board – Analog Devices Summary Sensor Fusion • • • • Algorithms for increased sensor performance Many different approaches Motion sensing is a highly pursued area Many companies offer products and tools