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.
- ROLE / AUTHORSHIP
- First author · independent study
- PUBLICATION STATUS
- Working paper · not peer reviewed
- MENTOR / AFFILIATION
- Faculty mentor — confirmation pending
Research question & abstract
A comparative study of classical and learning-based navigation methods under changing obstacle configurations, with a focus on safe and computationally efficient deployment.
Methods & evaluation
Implemented A*, RRT and a learned local planner; evaluated success rate, collision rate and runtime across 50 simulated and controlled physical trials.
System architecture
- 01
Obstacle configurations
- 02
classical and learned planners
- 03
simulation and physical trials
- 04
success, collision, and runtime comparison.
Results & findings
Sample finding: hybrid planning improved completion rates while maintaining a bounded computation budget.
Limitations & pending evidence
Sample study only; raw measurements, planner configurations, uncertainty estimates, and mentor confirmation are pending.