Thesis Examination Committee
Prof Shaojie SHEN, ECE/HKUST (Chairperson)
Prof Zexiang LI, ECE/HKUST (Thesis Supervisor)
Prof Fu ZHANG, ECE/HKUST
Local visual tracking, gimbal control, and an indoor global localization on a micro unmanned ground vehicle(UGV) are studied. In the robotics competition setting, a fast response dual-axial gimbal will results a higher shooting accuracy in an one-on-one combat, thus increasing the survival rate. The global localization gives a stable position estimation, and give directions to automate routine works to save game time.
In the local visual tracking, we mainly focus on solving the delay and outlier from the visual estimator and design the controller respectively. A constant velocity model on a linear Kalman filter predicts the movement of the target, and the angular velocity compensates the internal movement. A chi-square test rejects the non-Gaussian visual outlier. The controlled system is evaluated through the open loop system identification, identifying the mechanical design error and help to tune the controller parameter. In the global localization, a filter-based method fuses the indoor ultra-wideband module with inertial and wheel odometry. The methods are implemented on electrical four-wheel drive vehicles, each equipped with a dual-axial servo gimbal to shoot projectiles. The vehicle performance demonstrates the effects of the proposed methods.