Problem Statement
The objective of my project was to find a path for the robot that considers the trajectories of the trailers and avoids collisions.
Path planning for just the robot
Path planning including trailer trajectories.
The objective of my project was to find a path for the robot that considers the trajectories of the trailers and avoids collisions.
Path planning for just the robot
Path planning including trailer trajectories.
I tested multiple approaches to solve the problem:
I explored both general and standard trailers to understand their kinematics and control processes.
Standard Trailer System
General Trailer System.
This approach involves targeting convergence and obstacle avoidance for a non-standard tractor-trailer robot using a geometric method.
The control inputs are designed based on numerical integration of nonlinear differential equations using a 4th order Runge Kutta method. Obstacles are defined as either lines or ellipses. To ensure safety, I enclosed the tractor and each trailer in the smallest possible circles and adjusted the control inputs accordingly.
This approach uses a Hybrid A* algorithm to find a path while checking for trailers. It evaluates all possible configurations and removes those that result in collisions.
The Hybrid A* algorithm was chosen for its ability to include trailers in the path planning process, unlike the general A* algorithm. However, this increases computational complexity with the number of trailers.
This approach uses a deterministic grid-based search with a cost function calculated based on the error between the truck and trailer wheels.
A potentiometer measures the angle between the truck and the trailer to determine the error for the controller.
I calculated the maximum possible turning radius for the system, considering the maximum offset that can safely maneuver turns for N trailers. However, this approach proved less effective as it didn't account for realistic offsets.
This approach is based off of recent research in the field : ICRA Paper. While this approach works , it is not online and requires very high computation time and resources as the number of trailers increase.
I simulated the system in ROS and Gazebo using two approaches:
This project allowed me to explore various path planning algorithms for a robot with N-trailers. Each approach had its strengths and weaknesses, providing valuable insights for future work. I plan to refine these approaches and address the identified issues to develop a robust solution for global path planning with multiple trailers.