The complexity of robotic system has increased significantly in recent years due to the demand for higher levels of intelligence and autonomy. Modern robotic systems must be capable of autonomously adapting to changes in their operating environments while maintaining their objectives - even in highly uncertain and unstructured environments. Such robotic systems must display the capability to learn from experience, adapt to the changing working conditions, and seamlessly integrate information to-and-from humans. The core interest of the CLeAR research group is to enhance the performance and autonomy of robots through safe learning, enabling them to adapt to varying working conditions. Our research interests center around real-time optimisation-based control and estimation methods as well as machine learning teachniques with a heavy emphasis on application to autonomous systems.
Modeling these turbulent aerodynamic effects is a cumbersome task, and the resulting model may be overly complex and computationally infeasible. Combining Gaussian Process (GP) regression models with a simple dynamic model of the system has demonstrated significant improvements in control performance. However, direct integration of the GP models to the MPC pipeline poses a significant computational burden to the optimization process. Therefore, we present an approach to separate the GP models from the MPC pipeline by computing the model corrections using reference trajectory and the current state measurements prior to the online MPC optimization.
Accurate state estimation of the system is paramount for precise trajectory control for agile quadrotors; however, the high level of aerodynamic forces experienced by the quadrotors during high-speed flights makes this task extremely challenging. These complex turbulent effects are difficult to model, and the unmodelled dynamics introduce inaccuracies in the state estimation. In this work, we propose a method to model these aerodynamic effects using Gaussian Processes, which we integrate into the MHE to achieve efficient and accurate state estimation with minimal computational burden.
The project aims to develop a comprehensive plan for deploying methane sensing, modeling, and data management processes. This initiative seeks to create an integrated monitoring platform capable of continuous methane monitoring and reporting, facilitating swift responses to detected emissions.
The number of possible configurations of reconfigurable autonomous vessels exponentially grows as the total number of vessels increases, which imposes a technical challenge in modeling and identification. In this work, we propose a framework consisting of a real-time parameter estimator and a feedback control strategy, which is capable of ensuring high-accurate path tracking for any feasible configuration of vessels. Through experiments on different configurations of connected-vessels, we demonstrate stability of our proposed approach and its effectiveness in high-accuracy in path tracking.
We e develop a coordinated robust control scheme for a reconfigurable multi-vessel platform. The platform consists of N propeller-driven vessels each of which is capable of latching to another vessel to form a rigid body of connected vessels. Through experiments we assess trajectory tracking and disturbance attenuation performance of the control scheme in various configurations of the platform. Experiment results yield that average position and orientation tracking error are approximately 0.09m and 3 degrees, and the maximum tracking error-to-disturbance ratio is 1.12.
A simple learning strategy for feedback linearization control (SL-FLC) of systems with uncertainties is developed to facilitate accurate tracking in unknown/uncertain environments. The SL-FLC utilizes desired closed-loop error dynamics of the nominal system, which is minimized via the gradient-descent method to find the adaptation rules for feedback controller gains and disturbance estimate in the feedback control law, and finds the global optimum point. The SL-FLC framework can ensure the desired closed-loop error dynamics in the presence of disturbances. The performance of the SL-FLC is experimentally validated for the position tracking of a 3D-printed tilt-rotor tricopter UAV.
Plant traits, such as emergence rate, biomass, vigor, and stand counting are measured manually. This is highly labor-intensive and prone to errors. The robot, termed TerraSentia, is designed to automate the measurement of plant traits for efficient phenotyping as an alternative to manual measurements. We present results of an extensive field-test study of developed machine vision algorithm that shows that (i) the robot can track paths precisely with less than 5 cm error so that the robot is less likely to damage plants, and (ii) the machine vision algorithm is robust against interferences from leaves and weeds, and the system has been verified in corn fields at the growth stage of V4, V6, VT, R2, and R6 from five different locations. The robot predictions agree well with the ground truth with countrobot= 0.96× counthuman+ 0.85 and correlation coefficient R= 0.96.
A nonlinear moving horizon estimator identifies key terrain parameters using on-board robot sensors, and a robust learning-based nonlinear model predictive controller is designed to establish an effective control law for the 3D printed field robot traveling on rough terrain. The framework is designed to ensure high-precision autonomous path tracking in the presence of unknown wheel-terrain interaction.
When model-based control structures have to deal with uncertain and varying process conditions, it is inevitable to use adaptive models. Real-time estimators allow to make these model adaptations through online parameter estimation. Nonlinear moving horizon estimation method has been chosen as a state and parameter estimation algorithm because it considers the state and parameter estimation within the same problem and allows to incorporate constraints both on states and parameters.
To automate the trajectory following problem of an autonomous tractor-trailer system and also increase its steering accuracy, a centralized nonlinear model predictive control (CeNMPC) approach has been used. A fast CeNMPC is combined with nonlinear moving horizon estimation (NMHE) to obtain accurate trajectory tracking of an autonomous tractor-trailer system under unknown and variable soil conditions.
Instead of modeling the interactions between the subsystems prior to the design of a model-based control, we develop a control algorithm which learns the interactions online from the measured feedback error. The proposed learning algorithm is tested on the trajectory tracking problem of an autonomous agricultural tractor-trailer system in the presence of various nonlinearities and uncertainties in real time.