Optimization of Collective Bayesian Decision-Making in a Swarm of Miniaturized Vibration-Sensing Robots

Thiemen Siemensma,Bahar Haghighat

2024-12-19

Abstract:Inspection of infrastructure using static sensor nodes has become a well established approach in recent decades. In this work, we present an experimental setup to address a binary inspection task using mobile sensor nodes. The objective is to identify the predominant tile type in a 1mx1m tiled surface composed of vibrating and non-vibrating tiles. A swarm of miniaturized robots, equipped with onboard IMUs for sensing and IR sensors for collision avoidance, performs the inspection. The decision-making approach leverages a Bayesian algorithm, updating robots' belief using inference. The original algorithm uses one of two information sharing strategies. We introduce a novel information sharing strategy, aiming to accelerate the decision-making. To optimize the algorithm parameters, we develop a simulation framework calibrated to our real-world setup in the high-fidelity Webots robotic simulator. We evaluate the three information sharing strategies through simulations and real-world experiments. Moreover, we test the effectiveness of our optimization by placing swarms with optimized and non-optimized parameters in increasingly complex environments with varied spatial correlation and fill ratios. Results show that our proposed information sharing strategy consistently outperforms previously established information-sharing strategies in decision time. Additionally, optimized parameters yield robust performance across different environments. Conversely, non-optimized parameters perform well in simpler scenarios but show reduced accuracy in complex settings.

Robotics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to accelerate the decision - making process and improve accuracy by optimizing the collective Bayesian decision - making algorithm of the micro - vibration - sensing robot swarm. Specifically, the research aims to determine the dominant tile type (vibrating or non - vibrating) in a 1m × 1m tile surface and improve the existing decision - making algorithm by introducing a new information - sharing strategy.

### Problem Background
Traditional infrastructure detection usually uses static sensor nodes, and this method is very mature. However, when it comes to complex environments, static sensors may have limitations. Therefore, researchers have proposed the method of using mobile sensor nodes (i.e., micro - robots) for dynamic detection. These robots are equipped with an inertial measurement unit (IMU) for sensing and an infrared (IR) sensor for obstacle avoidance, and can perform efficient detection in a wider area.

### Research Objectives
The main objective of this paper is to solve a binary detection task through experimental settings, that is, to identify the main tile type (vibrating or non - vibrating) on a 1m × 1m tile surface. To achieve this goal, the author:
- Introduced a new information - sharing strategy (soft feedback) to accelerate the decision - making process.
- Developed a high - fidelity Webots robot simulator framework for optimizing algorithm parameters.
- Evaluated the performance of three information - sharing strategies (no feedback, positive feedback, and soft feedback) through simulation and actual experiments.

### Main Challenges
- **Decision - making Time**: When the filling ratio is close to 50%, the decision becomes particularly difficult because it is the most difficult to distinguish between vibrating and non - vibrating tiles at this time.
- **Accuracy**: Ensure high accuracy in environments with different levels of complexity.
- **Information - sharing Strategy**: Select an appropriate information - sharing strategy to accelerate decision - making and improve accuracy.

### Solutions
The author proposed a collective decision - making algorithm based on Bayesian inference and introduced three information - sharing strategies:
1. **No Feedback (No Feedback, \( u^- \))**: Each robot only broadcasts its latest observation results.
2. **Positive Feedback (Positive Feedback, \( u^+ \))**: Once a robot makes a final decision, it will broadcast this decision.
3. **Soft Feedback (Soft Feedback, \( u_s \))**: Combine long - term beliefs and current observation results to gradually guide the group to reach a consensus.

### Experimental and Simulation Settings
- **Experimental Platform**: A group of 3 - centimeter - sized micro - vibration - sensing robots were used and tested on 25 metal tile surfaces of 1m × 1m.
- **Simulation Environment**: Modeled in the Webots simulator to simulate the behavior of real robots, and noise calibration was carried out to reduce the gap between simulation and reality.

### Results
The experimental results show that the proposed soft - feedback strategy is significantly superior to the other two strategies in terms of decision - making time, and after optimizing the parameters, it can maintain stable performance in environments with different levels of complexity.

Through these improvements, this research shows the potential of micro - robot swarms in vibration - sensing infrastructure detection and provides valuable references for future related research.