Improved robot learning through a combination of machine learning for decision-making and water analysis

This study was published in Robot Learning It focuses on water analysis using a combination of decision-making and machine learning from recently developed robotic systems. The unique procedures applied by researchers have a major impact on improving the performance of robots that can detect, analyze and distinguish drinking water from Earth and other planets.

Robotic learning is a critical ability required for water analysis without human intervention. To this end, it is based on the application of leverage skills and training that allows the robot to learn the appropriate tasks and perform the activities efficiently. The advantages of autonomous robots that can perform water analysis are their rapid response in crisis situations, sustainable resource management, planetary exploration and reducing human intervention. Robot learning has been investigated for the development of various robotic tasks such as object manipulation, item cleaning, and interactive or multitasking learning, but water analysis using a combination of decision-making and machine learning (ML) has not been investigated.

Drinking water detection and distinction are important robotic tasks. Heavy metals and organic materials are toxic water pollutants that have caused health and environmental problems around the world. It is necessary to develop robots that can detect these contaminants and distinguish drinking water from Earth and other planets without human intervention.

For years, ML has been investigated without being combined with decision-making for water analysis. However, human decisions based on classification are a preliminary step in learning. Therefore, a combination of both processes is required for better analysis of robotics water samples. Therefore, researchers in the current research are applying a combination of decision-making and ML for improved robotic learning for water analysis.

Researchers applied a combination of decision making using Python's Microsoft Visual Studio code, using ideal solution (Topsis) and ML, using prioritization techniques. A Random Forest Classifier, a monitored ML algorithm, was used for water analysis.

Information about over 3200 water samples available in the Datasets section of the Kaggle website. “Water Quality and Portable Data Set” was used for water analysis. Dataset preprocessing was performed by completing the data table before analysis.

Topcys analysis of water samples showed that candidates with high value for profit criteria and low value for cost criteria had better rank. The same results were obtained through analysis of physicochemical properties and components of water samples. The ML simulation showed that using the modified code improved learning accuracy to 69%, and improved to 73% after using synthetic minority oversampling technology (SMOTE) for class balance and hyperparameter adjustment.

Robotic systems designed and developed for application of simulation software include electronic devices such as DC thrusters or drive motors, batteries, solar panels, and DC/DC converters. In this system, the control was run through a remote control and a command receiver. The remote control with four channels showed an adjustable speed. Additionally, you can adjust the direction well to make the model straight. At the input end the receive board exhibits reverse connection protection, with a self-healing fuse at the output, with adaptive and stable signal.

The designed robotic system allows water samples to be deposited in storage areas collected by steering the ship through the arms. The equipment required to build a ship's platform, including motors, batteries, solar panels, DC/DC converters, robotic arms and sensors, was developed to monitor physical and chemical parameters of water. After collecting water samples using an electric platform and a robotic arm. The sensor measures the physicochemical parameters of a sample in real time. Measurements for these sensors are handled by an onboard system. Here we combine a trained ML model with a decision-making process to classify water samples as easy to drink or unlinkable. Classification results guide the robot's decision-making process for storing and disposing samples. Therefore, hardware control is integrated with intelligent analysis.

To address the growing need for efficient water resource management and exploration, advanced robotic systems are equipped with autonomous water sample analysis capabilities. These enhancements are achieved through the integration of cutting-edge technologies into robotics, sensor systems, and artificial intelligence. Challenges related to sensor accuracy, data noise, and system scalability provide a balanced perspective on the feasibility of developed robotic water analysis systems by highlighting both potential limitations and strategies to overcome them. Therefore, addressing these issues ensures that the system remains practical and effective for real applications.

The results obtained are promising for the future of robotics, but further investigation is required to implement the results of current work in developed robotic systems, including the application of sensors. This will help develop a unique robotics platform for detection, analysis and distinction of drinking water on Earth and other planets.

Javabakht T, Pajaziti A, Buza S. Combination of decision-making and machine learning for improved robot learning for water analysis. Robot learning. 2025 (2): 0006, https://doi.org/10.55092/RL20250006