Visual-Inertial Localization for Skid-Steering Robots with Kinematic Constraints

Abstract

While visual localization or SLAM has witnessed great progress in past decades, when deploying it on a mobile robot in practice, few works have explicitly considered the kinematic (or dynamic) constraints of the real robotic system when designing state estimators. To promote the practical deployment of current state-of-the-art visual-inertial localization algorithms, in this work we propose a low-cost kinematics-constrained localization system particularly for a skid-steering mobile robot. In particular, we derive in a principle way the robot's kinematic constraints based on the instantaneous centers of rotation (ICR) model and integrate them in a tightly-coupled manner into the sliding-window bundle adjustment (BA)-based visual-inertial estimator. Because the ICR model parameters are time-varying due to, for example, track-to-terrain interaction and terrain roughness, we estimate these kinematic parameters online along with the navigation state. To this end, we perform in-depth the observability analysis and identify motion conditions under which the state/parameter estimation is viable. The proposed kinematics-constrained visual-inertial localization system has been validated extensively in different terrain scenarios.

Visual-based Lifelong Kinematics and Pose Estimation for Skid-Steering Robots

“…Alternatively, observability properties can also be directly analyzed under continuous-time nonlinear format, and a representative family of methods is to compute continuous-time observability matrices via selecting proper equations of Lie derivatives (Guo and Roumeliotis, 2013;Yang and Huang, 2019). Another category of widely used tools for observability analysis is to directly investigate the sensors' measurement models and robot's kinematic equations, to identify whether there exist sets of different states that are able to generate identical sensor measurements under the same system inputs, following the definition of observability (Jones and Soatto, 2011;Censi et al, 2013;Li and Mourikis, 2014b;Zuo et al, 2019b).…”

“…Our previous work (Zuo et al, 2019b)) performed observability analysis of localizing steering skid robot by using a monocular camera, wheel encoders, and an IMU, and showed that the skid-steering parameters are generally observable. In this work, we significantly extend the analysis in (Zuo et al, 2019b), by explicitly identifying the identifiable and nonidentifiable parameters with and without using the IMU.…”

“…Our previous work (Zuo et al, 2019b)) performed observability analysis of localizing steering skid robot by using a monocular camera, wheel encoders, and an IMU, and showed that the skid-steering parameters are generally observable. In this work, we significantly extend the analysis in (Zuo et al, 2019b), by explicitly identifying the identifiable and nonidentifiable parameters with and without using the IMU. For example, we show that, when an IMU is not included in the sensor system, it is still possible to design an estimator by modeling a reduced skid-steering intrinsic parameter vector and pre-approximating the rest.…”

Advancing robust state estimation of wheeled robots in degenerate environments: harnessing ground manifold and motion states

“…Moreover, non-holonomic constraints of vehicles have proven valuable in improving estimation accuracy [27]. Zuo et al derived kinematic constraints of vehicles using the instantaneous center of rotation and seamlessly integrated this information into a visual-inertial estimator, yielding superior results compared to conventional VIO systems [28]. Expanding on this work, Huang enhanced the robustness of odometry systems by incorporating non-holonomic constraints specific to vehicles.…”

Analysis of Tread ICRs for Wheeled Skid-Steer Vehicles on Inclined Terrain

“…Thus, constant approximations of tread ICR values for flat terrain under low inertia can be identified with offline experiments [16] [21] or online during navigation [22]. ICR kinematics has been employed as the basis for path tracking [23], for online power consumption estimation [24], for dynamic modeling [25], for online terrain classification [26] and, also, as constraints for simultaneous localization and mapping [27].…”