The Amazon River basin, home to one of the world’s largest freshwater reserves and unparalleled biodiversity, silently suffers from a vast environmental disaster caused by illegal mining, during which mercury is discharged into its waters. This contamination threatens aquatic ecosystems and poses serious risks to highly vulnerable populations. In response, this paper provides clues for a resilient and scalable system architecture for real-time water quality monitoring, tailored to the environmental and infrastructural challenges of the Amazon region. A detailed systematic literature review assesses state-of-the-art Internet of Things (IoT)-based monitoring techniques, focusing on key variables such as mercury concentration, temperature, turbidity, and pH. Special emphasis is given on communication technologies suitable for diverse settings—from Wi-Fi-enabled urban areas to remote rainforest regions where LoRa, NB-IoT, and WiLD (Wi-Fi over Long Distance) present viable alternatives. The study also highlights the role of the application layer in enabling data analysis, real-time alerts, and remote visualization of environmental conditions. This research contributes to developing time-efficient and sustainable monitoring strategies to support public health initiatives and ecological conservation by bridging technological innovation with the urgent environmental needs of one of the planet’s most critical biomes.

This work introduces a neuromorphic joint-space controller for the 4-DoF Serv-Arm robotic manipulator, implemented via a Spiking Neural Network (SNN) on low-cost embedded hardware. Unlike traditional methods based on inverse kinematics or Cartesian-space control, the proposed architecture operates directly on servo joint angles, receiving current and target configurations to predict the following joint-space action. A synthetic dataset with over 500,000 samples was generated to comprehensively cover the robot’s workspace, including oversampling of mechanically challenging configurations. The SNN was trained with surrogate gradients (snnTorch), and a systematic grid search optimized the hidden layer size, learning rate, and membrane decay factor. The final model, featuring a 128-neuron LIF hidden layer, achieved a Mean Absolute Error (MAE) of 11.06° on the validation set and generated smooth, reproducible trajectories during real-robot execution. Additionally, this study includes an empirical energy comparison between the SNN controller and an ANN baseline on a Raspberry Pi under identical motion routines. The results indicate that both models consume similar average powers, with the SNN showing a modest increase due to the temporal simulation overhead typical of spiking models on standard CPUs.