2-Degree-of-freedom Robot Path Planning using Cooperative Neural Fields

Area of Science:

• Robotics
• Artificial Intelligence
• Computational Neuroscience

Background:

• Path planning is a critical challenge in robotics, requiring efficient algorithms for robot navigation.
• Traditional methods often struggle with complex environments or real-time constraints.
• Dynamic programming offers a theoretical framework but can be computationally intensive.

Purpose of the Study:

• To propose a novel neural network-based approach for path planning in two degree-of-freedom (DOF) robots.
• To demonstrate the efficacy of using oscillatory neural network behavior for solving path planning problems.
• To validate the network's ability to find variational solutions to Bellman's dynamic programming equation.

Main Methods:

• A two-dimensional sheet of neurons representing the robot's workspace.
• Implementation of cooperative lateral interconnections between neurons to induce oscillatory behavior.
• Utilizing the network's dynamics to solve the path-planning problem.

Main Results:

• The proposed neural network successfully generates path plans for two DOF robots.
• The network exhibits oscillatory behavior that directly relates to solving the path-planning task.
• The obtained solutions are shown to be the variational solutions of Bellman's dynamic programming equation.

Conclusions:

• Neural networks offer a viable and potentially efficient alternative for robot path planning.
• The proposed oscillatory neural network architecture effectively addresses complex path planning challenges.
• This approach provides a new method for solving dynamic programming problems in robotics.