Robust actor-critic trajectory optimization and Autonomous harvesting robot integration
Saved in:
| Title: | Robust actor-critic trajectory optimization and Autonomous harvesting robot integration |
|---|---|
| Authors: | Tituana, Luis R |
| Committee Members: | Xu, Yunjun |
| Summary: | This dissertation is organized into two parts. First, we develop a neural network–based constrained trajectory optimization algorithm with stability guarantees, along with a neural network–driven numerical error correction. The framework reformulates the structure of an optimal control problem as a neural network optimization problem, where the solution space lies on a subspace manifold generated by a bio-inspired motion rule that produces open-loop control commands. To address changing conditions, real-world disturbances, and numerical errors, we propose an Actor–Critic–like architecture. In this setup, the Actor network outputs the optimal open-loop control for the optimized trajectory, while the Critic network compensates for disturbances and numerical errors. This configuration is embedded within a Receding Horizon Control (RHC) framework. Further, the robustness of the RHC is enhanced with the derivation of stability conditions in the form of bounds for the disturbances associated with numerical errors and changes in the environment. The proposed algorithms are validated through simulation experiments on nonlinear systems. The second part of the dissertation focuses on the integration and testing of strawberry harvesting robots operating in open-field farms. The development of such robotic harvesters directly addresses the growing shortage of agricultural labor in the United States. |
| URL: | https://stars.library.ucf.edu/etd2024/506 |
| Database: | OpenDissertations |
| Abstract: | This dissertation is organized into two parts. First, we develop a neural network–based constrained trajectory optimization algorithm with stability guarantees, along with a neural network–driven numerical error correction. The framework reformulates the structure of an optimal control problem as a neural network optimization problem, where the solution space lies on a subspace manifold generated by a bio-inspired motion rule that produces open-loop control commands. To address changing conditions, real-world disturbances, and numerical errors, we propose an Actor–Critic–like architecture. In this setup, the Actor network outputs the optimal open-loop control for the optimized trajectory, while the Critic network compensates for disturbances and numerical errors. This configuration is embedded within a Receding Horizon Control (RHC) framework. Further, the robustness of the RHC is enhanced with the derivation of stability conditions in the form of bounds for the disturbances associated with numerical errors and changes in the environment. The proposed algorithms are validated through simulation experiments on nonlinear systems. The second part of the dissertation focuses on the integration and testing of strawberry harvesting robots operating in open-field farms. The development of such robotic harvesters directly addresses the growing shortage of agricultural labor in the United States. |
|---|
