Abstract:Mobile robots are used in various fields, from deliveries to search and rescue applications. Different types of sensors are mounted on the robot to provide accurate navigation and, thus, allow successful completion of its task. In real-world scenarios, due to environmental constraints, the robot frequently relies only on its inertial sensors. Therefore, due to noises and other error terms associated with the inertial readings, the navigation solution drifts in time. To mitigate the inertial solution drift, we propose the MoRPINet framework consisting of a neural network to regress the robot's travelled distance. To this end, we require the mobile robot to maneuver in a snake-like slithering motion to encourage nonlinear behavior. MoRPINet was evaluated using a dataset of 290 minutes of inertial recordings during field experiments and showed an improvement of 33\% in the positioning error over other state-of-the-art methods for pure inertial navigation.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: in the case of pure inertial navigation, how to improve the positioning accuracy of mobile robots. Specifically, due to environmental limitations, mobile robots can often only rely on their inertial sensors (such as accelerometers and gyroscopes), and the readings of these sensors have noise and other errors, resulting in the navigation solution drifting over time. To solve this problem, the author proposes a new framework named MoRPINet, which uses deep learning to regress the distance traveled by the robot and enhances non - linear behavior through a serpentine motion pattern, thereby improving positioning accuracy.
### Problem Background
1. **Limitations of Inertial Sensors**
- Although inertial sensors (IMU) are simple to install, small in size, low in cost, and have a high sampling rate, when integrating over a long time, due to the influence of noise and other error terms, the navigation solution will gradually drift.
2. **Deficiencies of Existing Methods**
- Existing inertial navigation systems (INS) and model - based methods (such as MoRPI) perform poorly under pure inertial navigation conditions, especially in cases of low - amplitude, slow - motion, and the effectiveness of these methods is more limited.
### Proposed Solution
1. **MoRPINet Framework**
- **Serpentine Motion Pattern**: By making the mobile robot perform a snake - like meandering motion, the signal - to - noise ratio (SNR) of the inertial sensor data is increased, thereby providing more abundant feature information.
- **Deep Learning Network**: Use a neural network (D - Net) to regress the distance traveled by the robot, and combine the Madgwick filter to estimate the heading angle, and finally achieve high - precision positioning updates.
2. **Experimental Verification**
- Evaluation was carried out using a 290 - minute real - world inertial sensor data set, and the results show that MoRPINet improves the positioning error by 33% compared to other state - of - the - art methods.
### Main Contributions
1. **Innovative Serpentine Motion Pattern**: By imitating the meandering motion of a snake, the SNR of the inertial sensor data is increased, enabling the deep - learning network to extract more useful features.
2. **Efficient Deep - Learning Network**: A lightweight neural network architecture (D - Net) is proposed, which can run in real - time on edge devices for distance regression.
3. **Significant Performance Improvement**: Under pure inertial navigation conditions, MoRPINet shows better positioning accuracy and a faster update frequency than traditional INS and MoRPI methods.
### Formula Summary
- **Position Update Formula**
\[
x_{k + 1}=x_k + s_{Dnet,k}\cos(\psi_k)
\]
\[
y_{k + 1}=y_k + s_{Dnet,k}\sin(\psi_k)
\]
where \( s_{Dnet,k} \) is the distance estimated by D - Net, and \( \psi_k \) is the heading angle.
- **Average Heading Angle Calculation**
\[
\psi_k=\frac{1}{W}\sum_{i = 0}^{W}\psi_{AHRS}(t_{k,i})
\]
- **Evaluation Metrics**
- Position Root Mean Square Error (PRMSE)
\[
PRMSE(x_i,\hat{x}_i)=\sqrt{\frac{1}{N}\sum_{i = 1}^{N}|x_i-\hat{x}_i|^2}
\]
- Position Mean Absolute Error (PMAE)
\[
PMAE(x_i,\hat{x}_i)=\frac{1}{N}\sum_{i = 1}^{N}|x_i-\hat{x}_i|
\]
Through the above methods, the paper successfully solves the problem of low positioning accuracy of mobile robots under pure inertial navigation conditions and shows significant performance improvement.