Date

2022-5

Author

Deveci, Tuvana Deniz

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

236
views

211
downloads

Cite This

In real-world applications of motion planning and navigation, it is crucial to have a robust and accurate control policy. To achieve robustness and accuracy, the selected policy should handle the uncertainty in the process, which may arise from the surrounding environment or the process itself. However, most methods ignore the effects of uncertainty and cause inadmissible results for several applications. This thesis brings a solution to the addressed problem by proposing a trajectory-free motion control method that includes both the uncertainties and the constraints. In this solution, we propose constrained Stochastic Model Predictive Control (SMPC) for feedback motion planning application based on a Sampling-Based Neighborhood Graph (SNG). SNG defines a collision-free area to navigate the robot by adopting a sampling-based approach. While allowing a trajectory-free motion control, this approach provides a faster application due to its sparsity. SMPC, a receding horizon control approach for motion control, works on a stochastic system model by modeling the uncertainty as probabilistic distributions. Instead of using constraints of system boundaries, it defines chance constraints for handling stochasticity. The simulations of the proposed solution in a 2-D environment gave promising results. We tested different receding horizon control policies paired with various sampling-based motion planning approaches in MATLAB. We compared the successes of the controller with and without constraints and under chance constraints with stochastic and deterministic system models under different noise levels. The results show that having a sparse graph for motion planning affects the application’s speed and computational cost while only adopting a stochastic approach can provide safe and accurate robot motion planning and navigation under uncertainty.

Subject Keywords

Collision avoidance, Sampling-based neighborhood graph(SNG), Stochasticity, Estimation, Navigation

URI

https://hdl.handle.net/11511/97887

Collections

Graduate School of Natural and Applied Sciences, Thesis