# A Bionic Goal-Oriented Path Planning Method Based on an Experience Map

## TL;DR

This paper introduces a brain-inspired path planning method for robots that uses experience maps and reinforcement learning to navigate complex environments more efficiently than traditional methods.

## Contribution

The novel contribution is a bionic path planning approach using NeuroSLAM and reinforcement learning to create adaptive and efficient global path planning for mobile robots.

## Key findings

- The proposed method outperforms Dijkstra’s algorithm in adaptability and computational efficiency.
- The experience map and predictive model enable accurate sequence prediction and goal-oriented navigation.
- Reinforcement learning enhances long-term reward estimation for optimal path planning.

## Abstract

Brain-inspired bionic navigation is a groundbreaking technological approach that emulates the biological navigation systems found in mammalian brains. This innovative method leverages experiences within cognitive space to plan global paths to targets, showcasing remarkable autonomy and adaptability to various environments. This work introduces a novel bionic, goal-oriented path planning approach for mobile robots. First, an experience map is constructed using NeuroSLAM, a bio-inspired simultaneous localization and mapping method. Based on this experience map, a successor representation model is then developed through reinforcement learning, and a goal-oriented predictive map is formulated to address long-term reward estimation challenges. By integrating goal-oriented rewards, the proposed algorithm efficiently plans optimal global paths in complex environments for mobile robots. Our experimental validation demonstrates the method’s effectiveness in experience sequence prediction and goal-oriented global path planning. The comparative results highlight its superior performance over traditional Dijkstra’s algorithm, particularly in terms of adaptability to environmental changes and computational efficiency in optimal global path generation.