Research Papers & Publications
Robot path planning and neural optimization study summaries, including methods, findings, limitations, and manuscript links. These are sample studies, not verified publications.
Learning-Based Path Planning in Dynamic Environments
A comparative study of classical and learning-based navigation methods under changing obstacle configurations, with a focus on safe and computationally efficient deployment.
METHODOLOGY
Implemented A*, RRT and a learned local planner; evaluated success rate, collision rate and runtime across 50 simulated and controlled physical trials.
FINDINGS & LIMITATIONS
Sample finding: hybrid planning improved completion rates while maintaining a bounded computation budget.
Neural Optimizers for Structured Objective Functions
Explores whether recurrent networks can learn update rules that generalize across families of differentiable objectives.
METHODOLOGY
Train an LSTM on synthetic quadratic functions and compare convergence against gradient descent and Adam on unseen condition numbers.
FINDINGS & LIMITATIONS
Sample finding: improved convergence within the training distribution; limited generalization to nonconvex objectives.