Abstract:This project has conducted research on robot path planning based on Visual SLAM. The main work of this project is as follows: (1) Construction of Visual SLAM system. Research has been conducted on the basic architecture of Visual SLAM. A Visual SLAM system is developed based on ORB-SLAM3 system, which can conduct dense point cloud mapping. (2) The map suitable for two-dimensional path planning is obtained through map conversion. This part converts the dense point cloud map obtained by Visual SLAM system into an octomap and then performs projection transformation to the grid map. The map conversion converts the dense point cloud map containing a large amount of redundant map information into an extremely lightweight grid map suitable for path planning. (3) Research on path planning algorithm based on reinforcement learning. This project has conducted experimental comparisons between the Q-learning algorithm, the DQN algorithm, and the SARSA algorithm, and found that DQN is the algorithm with the fastest convergence and best performance in high-dimensional complex environments. This project has conducted experimental verification of the Visual SLAM system in a simulation environment. The experimental results obtained based on open-source dataset and self-made dataset prove the feasibility and effectiveness of the designed Visual SLAM system. At the same time, this project has also conducted comparative experiments on the three reinforcement learning algorithms under the same experimental condition to obtain the optimal algorithm under the experimental condition.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper primarily investigates the problem of robot path planning based on visual SLAM (Simultaneous Localization and Mapping). Specifically, the paper aims to address the following aspects: 1. **Visual SLAM System Construction**: - Study the basic architecture of visual SLAM and develop a visual SLAM system capable of dense point cloud mapping based on the ORB-SLAM3 system. 2. **Map Conversion**: - Convert dense point cloud maps into maps suitable for 2D path planning. The specific steps include converting the dense point cloud map into an octomap, and then performing a projection transformation to generate a grid map. This process transforms the dense point cloud map, which contains a lot of redundant information, into a lightweight grid map suitable for path planning. 3. **Reinforcement Learning-based Path Planning Algorithm**: - Compare three reinforcement learning algorithms: Q-learning, DQN, and SARSA. Experimental verification shows that DQN has the fastest convergence speed and the best performance in high-dimensional complex environments. 4. **System Verification**: - Conduct experimental verification of the visual SLAM system in a simulation environment. Using open-source datasets and self-made datasets, the feasibility and effectiveness of the designed visual SLAM system are demonstrated. Additionally, the three reinforcement learning algorithms are compared under the same experimental conditions to obtain the optimal algorithm. In summary, the core objective of this paper is to achieve robot localization and map construction using visual sensors, and based on this, obtain the optimal path for mobile robots. By constructing a visual SLAM system, performing map conversion, and studying reinforcement learning-based path planning algorithms, the paper aims to enhance the autonomous navigation capability of robots in unknown environments.

Research on Robot Path Planning Based on Reinforcement Learning

Visual Local Path Planning Based on Deep Reinforcement Learning

Research on path planning algorithm of mobile robot based on reinforcement learning

Reinforcement Learning-Based Path Planning Algorithm for Mobile Robots

SLAM, Path Planning Algorithm and Application Research of an Indoor Substation Wheeled Robot Navigation System

A Mapless Local Path Planning Approach Using Deep Reinforcement Learning Framework

Deep Reinforcement Learning for Mobile Robot Path Planning

Path Planning for Active V-Slam Based on Reinforcement Learning

Research on AGV Path Planning Based on Reinforcement Learning

Reinforcement Learning Path Planning Method with Error Estimation

Autonomous Navigation by Mobile Robot with Sensor Fusion Based on Deep Reinforcement Learning

Deep Reinforcement Learning for Indoor Mobile Robot Path Planning

Path Planning of Autonomous Mobile Robot in Comprehensive Unknown Environment Using Deep Reinforcement Learning

On Reward Shaping for Mobile Robot Navigation: A Reinforcement Learning and SLAM Based Approach

Path Planning for Autonomous Vehicles in Unknown Dynamic Environment Based on Deep Reinforcement Learning

A Review of Mobile Robot Path Planning Based on Deep Reinforcement Learning Algorithm

Improved Robot Path Planning Method Based on Deep Reinforcement Learning

A Method to Plan the Path of a Robot Utilizing Deep Reinforcement Learning and Multi-Sensory Information Fusion

A Novel Autonomous Exploration Algorithm Via LiDAR/IMU SLAM and Hierarchical Subsystem for Mobile Robot in Unknown Indoor Environments

A Distributed Path Planning Algorithm via Reinforcement Learning

Research on path planning algorithm of unmanned ground platform based on reinforcement learning