14 papers accepted for presentation.
The beside an author opens their certificate of presentation, which anyone can verify.
Arianne P EspinosaCertificate,Celine Kirsten C MercadoCertificate,Bernize Lexine M AngelesCertificate
The Philippines faces critical challenges in its sewage system, driven by bacterial contamination, chemical contamination, and the accumulation of decaying organic matter. These conditions pose health risks through the emission of harmful gases, such as hydrogen sulfide, methane, ammonia, and carbon dioxide, produced by decaying waste. Moreover, everyday items such as grease, paper towels, and hygiene products are continuously being flushed down drains, clogging the sewer line and further increasing the risk of flooding in these areas. While established solutions focus on wastewater treatment, the researchers propose a different approach—one that focuses on detection to prevent further contamination. SynchroStream is an upgrade to the average sewage system, using four types of sensors to monitor water levels and close grates in the presence of waste, as well as to send data to local authorities via LoRa radio. It utilizes ultrasonic sensors to detect the current water level and debris found in the water. At the same time, a layer of EPDM (ethylene propylene diene monomer) rubber is combined with servo motors to prevent waste from entering the openings. Upon detecting rain using its BMP280 and DHT22, the motors move the rubber to collect rainwater. Additionally, a separate DHT22 sensor and waterflow sensor provide early detection of mosquitoes and notify local government units with updated data about whether there is a need to sanitize and perform maintenance on the area using a machine learning-based predictive maintenance utilizing a Random Forest classification model over the LoRa radio via the ESP32 Development Boards. Then, the data from each sensor is collected by the first ESP32 board and dispatched to the LoRa transmitter. Once it reaches the LoRa receiver, the data is processed by a second ESP32 before being displayed on the dashboard. This project aims to provide an efficient, cost-effective system to mitigate flooding in urban areas and reduce the exposure of the public to toxic gases and diseases associated with them.
Robot Hardware and Software Design of intelligent robots,Robot Cognition and Learning,High-speed object tracking
Walter Iran Amador ValenzuelaCertificate,Bo ZhouCertificate,Meng Cheng LauCertificate
Autonomous driving remains a challenging problem in mobile robotics due to the need for reliable perception, decision making, and control under limited computational resources. This paper presents a lightweight hybrid vision framework for autonomous driving in the FIRA Urban Challenge using the AGILEX LIMO platform. The proposed system combines classical computer vision techniques for lane and obstacle detection with a YOLOv8n-based traffic sign recognition module. A hierarchical finite state machine architecture coordinates line following, junction handling, traffic sign interpretation, and obstacle avoidance, while a proportional-derivative controller generates real-time steering commands. To evaluate the suitability of learning-based perception for embedded robotic platforms, a comparative study between AprilTag and YOLO-based traffic sign recognition was conducted. Experimental results demonstrate that although AprilTags provide superior geometric accuracy, the YOLO-based approach offers greater robustness under partial occlusion, varying illumination conditions, and non-ideal viewing angles. Operating entirely on onboard CPU resources using only the RGB stream from an RGB-D camera, the proposed framework achieved an average processing rate of 14.5 FPS while successfully performing autonomous navigation tasks in FIRA-style environments. The results demonstrate that combining lightweight deep learning with classical vision methods provides an effective solution for embedded autonomous driving applications.
Amirmohammad Zarif Shahsavan NejadCertificate,Helia MokhtariCertificate,Azita ShirazipourCertificate,Soroush SadeghnejadCertificate,Amirmahdi ZarifCertificate
The integration of artificial intelligence (AI) and natural language processing (NLP) in healthcare has demonstrated transformative potential for disease diagnosis and clinical decision support. This paper presents a comprehensive analysis of current AI and NLP methodologies in healthcare, examining their applications across medical imaging, clinical text analysis, and diagnostic decision support systems. We systematically evaluate performance metrics from 50+ peer-reviewed studies, revealing that AI systems achieve diagnostic accuracies ranging from 88.7% to 96.2% across different medical specialties, often matching or exceeding human expert performance. Our analysis identifies key implementation challenges including data quality, algorithmic transparency, and clinical workflow integration. We propose a framework for responsible AI deployment in healthcare settings and discuss emerging trends in multimodal diagnostic approaches. The findings suggest that while AI and NLP technologies offer significant clinical benefits, successful implementation requires careful attention to technical validation, ethical considerations, and stakeholder engagement.
Amirmahdi ZarifCertificate,Amirmohammad Zarif Shahsavan NejadCertificate,Jacky BaltesCertificate,Kuo-Yang TuCertificate,Soroush SadeghnejadCertificate
This study presents a comprehensive Open Innovation model, infused with Systems Thinking principles, to investigate the FIRA Innovation & Business League’s role as a global platform for propelling robotics and emerging technologies forward. By framing the league as a multi-stakeholder ecosystem, the research delineates a layered framework that elucidates interactions among startups, established organizations, and the wider community. Employing a mixed-methods approach, the study illuminates how the league enables knowledge exchange, technology transfer, and entrepreneurial collaboration. The findings underscore that this integrated model accelerates innovation cycles, bolsters startup performance, and cultivates robust linkages among industry, academia, and government. A SWOT analysis further unveils the challenges and opportunities inherent in maintaining open innovation amid a swiftly evolving technological terrain. The paper culminates in underscoring the pivotal role of problem-driven innovation formats and sustainable ecosystem architectures, with special attention to the league’s forthcoming iterations, including its anticipated 2026 expansion into Canada.
台師大陳仲崴Certificate,LIU, CHIA-HENGCertificate
This paper presents a fully onboard embedded autonomous race track system targeting the FIRA Challenge – Autonomous Cars (Pro) Race Track task. The system is built on a Carson T410R 1:10 scale RC car equipped with a Raspberry Pi 4B single-board computer, a monocular camera, and a BMI160 IMU. A Fast-SCNN semantic segmentation network fine-tuned on 53,000 race-track frames achieves mIoU = 0.732 and maintains approximately 56–58 FPS under full ROS 2 node load on the Raspberry Pi 4B. A segmentation-to-path pipeline converts dashed-line masks into navigation target points via connected-component centroid extraction and linear polynomial fitting. An IMU gyroscope feedforward term pre-corrects camera servo angle to compensate for the ~40 ms visual latency inherent to the embedded pipeline. A raceoriented control logic integrates nonlinear camera-to-wheel steering, rate-limited wheel commands, line-loss stopping, obstacle-triggered steering offsets, curvature-based speed limiting, lateral-G throttle reduction, and IMU-failure safe mode. Kinematic simulation on an elliptical track yields mean lateral error 0.67 cm (max 2.66 cm). Practice-track testing shows approximately 90% corner success with autonomous line-recovery capability. The system operates entirely onboard with no wireless control links, satisfying FIRA Rule 2.8 and qualifying for the onboard-processing autonomy bonus (Ka=1.0).
autonomous racing, fira challenge, semantic segmentation, pure pursuit, imu compensation, ros 2, embedded systems
CHENG-JIE PENGCertificate,Yun-Tzer ChenCertificate
This study uses the Darwin OP3 humanoid robot as the experimental platform to implement various challenging tasks and athletic movements, such as weightlifting, archery, and the Spartan Race events in the FIRA competition. By applying a multi-node communication architecture under the ROS environment, image recognition, proportional control, motor control, and object detection models, the stability and accuracy of the robot during operation are improved.
Brooke Zita VrbanicCertificate,Meng Cheng LauCertificate
This paper presents the development and evaluation of an autonomous archery system for humanoid robots. The purpose of the project was to evaluate how effectively a humanoid robot can perform a precision sports task that requires coordinated biomechanics, reliable perception, and consistent autonomous decision-making. Archery was selected because it integrates key components of humanoid robotics, such as kinematics, balance, vision, and timing, within a single repeatable activity. The robot operated in ROS1 using Dynamixel actuators and a custom bow and arrow setup. A YOLOv9 vision system was implemented to detect the target in real time and provide accurate positions. The logic combined perception, planned alignment, predefined joint motion sequences, and time-based synchronisation for arrow release. All experiments were carried out physically in a controlled indoor environment. The robot, Polaris, was tested multiple times at varying target rotation speeds, and performance was evaluated through accuracy, repeatability, latency, and overall stability of the system. The results showed that Polaris could consistently hit the target at moderate rotation speeds, with performance decreasing as the target speed increased. The system demonstrated reliable integration of perception and motion control, with successful hits achieved. Key limitations included occasional actuator torque issues, vision latency after extended operation, and sensitivity to battery voltage during movement. Despite these constraints, the project achieved its primary objective of demonstrating that a humanoid robot can autonomously perform archery with measurable accuracy using lightweight control methods and realtime visual feedback.
Chih-Cheng Liu 劉智誠Certificate,JAESIK JEONGCertificate,CHENG LIN KUOCertificate
This study investigates deep reinforcement learning for bipedal robot gait control using NVIDIA Isaac Gym. A high-degree-of-freedom simulation is developed, and the agent first learns under an unconstrained baseline by directly controlling joint angles. To improve stability and realism, inverse kinematics and Zero Moment Point constraints are progressively introduced, ensuring feasible motion and center of mass balance within the support polygon. Some experiments further employ reference foot trajectories to guide learning and enhance efficiency. The effects of different control conditions—guided versus unguided and with or without constraints—are analyzed to understand how knowledge-based limitations influence the agent’s learning performance and gait behavior.
deep reinforcement learning,proximal policy optimization,bipedal locomotion,physics simulation,robot simulator,isaac gym,Robot Cognition and Learning
Armin Baratian SarabiCertificate,Amin AramiCertificate,Amirmohammad Zarif Shahsavan NejadCertificate,Amirmahdi ZarifCertificate,Soroush SadeghnejadCertificate
The development of humanoid robots, particularly in enhancing human-robot interaction, is a rapidly advancing field. This paper explores the simulation of the Thormang humanoid robot and the development of advanced grasping models using machine learning techniques. Our research delves into the intricacies of kinematics, dynamics, and computer vision, paving the way for more intuitive and effective humanrobot interactions. Using and comparing different 6D pose models such as CASAPose, we continue to perform the provided tasks in object manipulation.
Shi-Han WangCertificate,Hanjaya MandalaCertificate,Saeed SaeedvandCertificate,Jacky BaltesCertificate
Self-balancing wheeled robots present unique challenges for mobile manipulation due to continuous variations in height, orientation, and tilt caused by the robot’s self-balancing dynamics. In this paper, we present a PPO-based framework with a five-stage dense reward function that shapes the policy to reach, grasp, lift, transport, and accurately place objects on our self-balancing two-wheeled robot under a generalized multiobject training setting, where all three objects (Mug, Drill, and Dumbbell) are trained simultaneously in a single policy. The fivestage reward guides exploration in this high-dimensional floatingbase manipulation task without requiring explicit kinematic programming. Our policy achieves a success rate of 69.78% on the Drill task and sub-centimeter placement accuracy of 9.52 mm. The experiments also reveal the spontaneous emergence of prealignment behavior, where the robot learns to reorient asymmetric objects into graspable poses without explicit instruction, suggesting that the reward structure encourages adaptive manipulation strategies beyond what was explicitly programmed.
Jee-Hyun YangCertificate,JAESIK JEONGCertificate
This paper presents heuristic design methodologies for humanoid robots developed through practical experience in international humanoid robot competitions. Since humanoid robot development requires close integration between design and engineering, this study proposes an approach that connects technical requirements with design-oriented decision-making. The proposed approach focuses on mechanical and electrical design strategies derived from repeated prototyping, testing, and competition-based refinement. For mechanical design, this paper introduces lightweight structural concepts and optimized wiring layouts using hollow shafts. These strategies improve robustness, reduce unnecessary weight, and support efficient dynamic movement in diverse environments. For electrical design, the paper explains component selection based on performance, reliability, and operational efficiency. Main and sub controllers, sensors, actuators, and communication modules are selected to support control algorithms, artificial intelligence (AI), real-time data processing, and stable communication. Based on practical experience, this paper discusses key challenges in humanoid robot development and practical solutions obtained through iterative improvement. By sharing design strategies grounded in real-world implementation, this research offers useful insights for developing effective humanoid robot platforms.
Robot Hardware and Software Design of intelligent robots
Arianne P EspinosaCertificate,Celine Kirsten C MercadoCertificate,Bernize Lexine M AngelesCertificate
The Philippines faces critical challenges in its sewage system, driven by bacterial contamination, chemical contamination, and the accumulation of decaying organic matter. These conditions pose health risks through the emission of harmful gases, such as hydrogen sulfide, methane, ammonia, and carbon dioxide, produced by decaying waste. Moreover, everyday items such as grease, paper towels, and hygiene products are continuously being flushed down drains, clogging the sewer line and further increasing the risk of flooding in these areas. While established solutions focus on wastewater treatment, the researchers propose a different approach—one that focuses on detection to prevent further contamination. SynchroStream is an upgrade to the average sewage system, using four types of sensors to monitor water levels and close grates in the presence of waste, as well as to send data to local authorities via LoRa radio. It utilizes ultrasonic sensors to detect the current water level and debris found in the water. At the same time, a layer of EPDM (ethylene propylene diene monomer) rubber is combined with servo motors to prevent waste from entering the openings. Upon detecting rain using its BMP280 and DHT22, the motors move the rubber to collect rainwater. Additionally, a separate DHT22 sensor and waterflow sensor provide early detection of mosquitoes and notify local government units with updated data about whether there is a need to sanitize and perform maintenance on the area using a machine learning-based predictive maintenance utilizing a Random Forest classification model over the LoRa radio via the ESP32 Development Boards. Then, the data from each sensor is collected by the first ESP32 board and dispatched to the LoRa transmitter. Once it reaches the LoRa receiver, the data is processed by a second ESP32 before being displayed on the dashboard. This project aims to provide an efficient, cost-effective system to mitigate flooding in urban areas and reduce the exposure of the public to toxic gases and diseases associated with them.
High-speed object tracking,Complex motion planning,Path Planning and Collision Avoidance in Autonomous Vehicles,Robot Cognition and Learning
Muhammad Haziq Zuhairy Bin Mohd HasmadiCertificate,HELMI BIN JAMALUDINCertificate,Najmi Hafizi Bin ZabawiCertificate,SHAHMIE HAZIQ BIN SAIFUL AZWACertificate
Safe and stable high-speed obstacle avoidance remains a critical challenge for autonomous mobile robots (AMRs) operating on constrained tracks. This paper presents a trajectory planning and sensor fusion framework designed for a 1/12-scale autonomous car robot. To prevent wheel slip and lateral instability during sudden avoidance maneuvers, we propose a trapezoidal steering and velocity profiling algorithm. The steering angle is adjusted through coordinated ramp, hold, and settle phases, while the vehicle decelerates to maintain traction and accelerates back to nominal speed upon maneuver completion. The system incorporates a dual-axle active steering geometry in crab mode, allowing lateral translation with minimal yaw deviation. Perception is achieved by fusing a low-latency ultrasonic sensor with a bird's-eye view camera pipeline. To prevent failure under sensor degradation, an area-based visual distance model is integrated as a real-time fallback. Hardware-in-the-loop experiments conducted on a physical vehicle prototype validate the effectiveness of the proposed architecture. The results show that the trapezoidal trajectory planner achieves a 100% success rate in avoiding obstacles at speed, reducing peak lateral deviation to 3.5 cm and maintaining track stability where baseline static-steering methods fail.
trajectory planning,trapezoidal velocity profile,sensor fusion,autonomous car robot,crab steering,dual ackerman.
Shubham VermaCertificate
This paper synthesizes three empirical studies on LLM agent security. First, we present a three-axis taxonomy (attack vector, mode of operation, goal) for prompt injection attacks, validated against 30+ attack patterns. Second, we analyze the Model Context Protocol (MCP) attack surface: 30 CVEs in 60 days across 53K+ skills, with 71% tool description poisoning and 38% of servers lacking authentication. Third, we introduce AegisCLI, a local-first DevSecOps orchestrator using CodeLlama-13B (Ollama), SARIF v2.1.0, OPA/Rego, and Redis to reduce tool-switching overhead by 62% and security debt by 43% across 500+ engineers. Combined, these contributions form an end-to-end agentic security framework spanning threat taxonomy, protocol-level vulnerability measurement, and privacypreserving remediation.