Tire Force Estimation of Wheeled Mobile Robots

We proposed a real-time tire forces and friction coefficient estimation technique using the constrained Kalman filter (TFF-CKF). The proposed TFF-CKF estimates the state and tire forces using onboard sensor measurements and selects the most proper friction coefficient using the Bayesian hypothesis selection algorithm. With the selected friction coefficient and tire model, the proposed TFF-CKF updates the tire forces using the constrained Kalman filter to enhance the robustness and accuracy. The proposed technique is verified by simulations and experiments, and the mean and standard deviation of the estimated friction coefficient are, respectively, 0.8307 (typical asphalt: 0.8-1.0) and 0.0198. (IROS2018)

Camera-GNSS-IMU Sensor Fusion

​GNSS-IMU based sensor fusion is widely used for autonomous flying, which yet suffers from the inaccuracy and drift of the GNSS signal and also the failure with the loss of GNSS (e.g., indoor flying). For this, we propose a new framework for camera-GNSS-IMU sensor fusion, which, by fusing monocular camera information with GNSS and IMU, can improve accuracy and robustness of the autonomous flying. For camera and GNSS-IMU calibration, a new Kalman filter is also proposed, which runs in parallel with state estimation EKF and also utilize multiple keyframes generated from the camera information. (DroneCAN2016)

SLAM-IMU Fusion for Indoor Flying

GNSS has been widely used for outdoor autonomous flying of drone, which yet becomes unavailable for indoor flying.  For this, we implement EKF-based sensor fusion by fusing mono-camera, rangefinder and IMU (i.e., gyroscope, accelerometer and magnetometer).  We particularly adopt ORB-SLAM for monocular camera information processing to achieve fast localization of the drone, which is crucial for autonomous flying control.  The drone is controlled by our own backstepping flying controller. (RAS2014

Formation Flying w/ GNSS-IMU Fusion

Outdoor formation flying of three custom-built drones in a circular shape. For this, we utilize EKF-based state estimation in SE(3), which uses the information of GNSS, IMU, and barometer, consisting of : (1) state propagation by using the gyroscope and accelerometer of the IMU; and (2) measurement update by using the compass, barometer and GNSS.  The drones are also controlled by in-house built backstepping control law presented in (RAS2014). 

Autonomous Dynamic Driving

Control framework to enable an nonholonomic WMR to autonomously drive fast enough, yet, still less than a certain threshold to prevent wheel slippage. For this, instead of Newtonian vehicle modeling, we adopt Lagrange-D’Alembert formulation to explicitly relate the system’s state/control with constraint force, so that we can predict/detect possibility of violating no-slip condition.  We also experimentally show that our control law, albeit based on simple Coulomb friction model, can still allow for dynamic driving of a RC car  with no slippage.  (ICRA2014)

© 2016 by INRoL
Proudly created with

1 Gwanak-Ro, Gwanak-Gu

(Office) Bldg 301 #211, #1203/1204

(Lab) Bldg 312 #202/203/204 

Seoul National University

Seoul, 08826, Korea

(Phone) +82-2-880-1690

(Email)         (Map)