Embodied intelligence and service robotics
Multimodal situational awareness, safe navigation, robot vision and systems integration, forming a complete perception-to-execution chain.
LINKINWISE / CREATOR
Dr. Jia Yin
Eighteen years of study and work abroad, combining AI, robotics, engineering management and cross-institution collaboration. During his doctorate he took part in Singapore's National Robotics Programme and the Toyota HSR service robot project, gaining experience in robot vision, systems integration, intelligent control and engineering deployment. He now focuses on embodied intelligence education, ROS2 curricula, AI literacy education, robotics competitions and university-enterprise joint AI and robotics laboratories.
Senior lecturer in embodied intelligence | University teacher in AI and robotics | Systems integration
PROFILE
Profile and development path
- Research and technology
- Embodied intelligence, service robotics, multimodal situational awareness, safe navigation, robot vision and systems integration.
- Current affiliation
- Huanghe University of Science and Technology, AI and robotics.
- Technical thread
- A complete chain of perception, understanding, decision and execution, applied to public services, building inspection, smart agriculture, disaster rescue, elderly care and engineering education.
Development path
Engineering foundation
Civil engineering with a minor in computer engineering
Site management
Organising a major rail transit project, quality and safety
Robotics research
Service robots, multimodal perception and safe navigation
University translation
Courses, projects, competitions, research and industry collaboration
Core capability structure
ROS2 engineering education
Turns Linux, ROS2 communication, URDF, Gazebo, RViz, Nav2, robot vision and real tasks into courses, labs and competition projects.
Engineering systems and project management
Requirement breakdown, solution design, version management, field testing, quality and safety, risk control and multi-party coordination.
Scenario translation and collaboration
Applied translation and technical collaboration across public services, building inspection, smart agriculture, disaster rescue, elderly care and engineering education.
EXPERIENCE
Career and education
Listed in reverse chronological order.
Huanghe University of Science and Technology
2025—present
University teaching, embodied intelligence education and engineering translation
- Building project-based courses in robotics fundamentals, ROS2, AI literacy and embodied intelligence.
- Organising Linux, ROS2 communication, URDF, Gazebo, RViz, Nav2, robot vision and real tasks into courses, labs and competitions.
- Connecting courses, projects, competitions, research and industry, moving from lecturing towards engineering practice.
Rosen Bridge Technology, Singapore
2023—2024
Robotics, systems integration and application translation
- Co-founded Rosen Bridge Technology in Singapore, advancing early work in robotics and systems integration.
- Took part in international industry exchange, focusing on productisation and partnership translation in real business scenarios.
- Built practical experience spanning technical proposals, partnership communication and project delivery.
Singapore University of Technology and Design (SUTD)
2018—2023
Doctoral research in service robotics and multimodal situational awareness
- Researched on the Toyota HSR human support robot and the BELUGA autonomous mobile cleaning robot.
- Hazard target recognition, RGB-D 3D localisation, acoustic event detection, sound source localisation, safe distance estimation, path planning and real-robot validation.
- Awarded the doctorate in Engineering Product Development in 2023, completing a chain from data collection and algorithm development to ROS integration and real-scene testing.
Land Transport Authority, Singapore
2014—2018
Site management and multi-party coordination for a major rail transit project
- Took part in site management, quality and safety, and multi-discipline coordination.
- Organised and coordinated a workforce of over 400 at peak, handling schedule, interfaces, risk and acceptance.
- Built project management capability: breaking complex tasks into requirements, design, versioning, testing, risk and milestone acceptance.
Nanyang Technological University (NTU), Singapore
2010—2014
Civil engineering major with a minor in computer engineering
- Structured engineering thinking through civil engineering; algorithms, software and systems through computer engineering.
- This cross-disciplinary background later grounded the combination of engineering scenarios, algorithms and robotics.
- Developed the ability to understand and coordinate across disciplines around real system problems.
ARICC Robotics Laboratory, Singapore Polytechnic
2010
RoboCup domestic service robot competition
- Competed in the RoboCup 2010 domestic service robot category, leading vision system development and optimisation.
- The team placed sixth among 26 international university teams.
- Early exposure to service robot perception, task execution and international competition teamwork.
WORK
Core work and collaboration
Representative work
Vision perception and autonomous operation on Toyota HSR
Tabletop waste recognition, RGB-D 3D localisation, cleaning path planning and door handle detection, with the mobile base and manipulator executing together — a closed loop from perception to operation.
Hazard recognition and safe navigation in complex public environments
For low-feature or occluded hazards such as escalators, stairs and glass doors, contextual, visual and depth information are fused for hazard recognition and safe distance estimation.
Acoustic hazard area perception when vision is limited
Sound event detection and source direction localisation identify hazard areas and drive coverage path replanning, so the robot actively avoids risk under low light and occlusion.
Project-based embodied intelligence and ROS2 engineering education
Organising knowledge into courses, labs, competitions and real projects, moving students from using tools to completing engineering tasks.
Collaboration directions
Embodied intelligence
Multimodal perception, short-horizon prediction, risk modelling and robot decision making.
Service robotics
Public services, elderly care, building inspection, smart agriculture and disaster rescue.
ROS2 course building
Curricula, lab projects, teacher training and engineering practice platforms.
Systems integration
Robot interfaces, edge controllers, navigation and real-robot validation.
Talent development
An integrated chain of courses, projects, competitions, research and industry.
Results translation
Scenario mapping, technical solution design and joint development.
RESEARCH
Research foundation and representative results
Research foundation
Long-term research in embodied intelligence, service robotics, robot vision, multimodal situational awareness, safe navigation and robot systems integration. He received his doctorate in Engineering Product Development from SUTD in 2023; his thesis concerned situational awareness and autonomous operation of cleaning robots in complex public environments.
During his doctorate he worked on the Toyota HSR human support robot and the BELUGA autonomous mobile cleaning robot: visual and RGB-D perception, recognition of low-feature and occluded hazards, acoustic event detection and source localisation, safe distance estimation, path planning and physical validation — a complete chain from data collection and algorithm development to ROS integration and real-scene testing.
Key doctoral research projects
National Robotics Programme · Robotics Enabling Capabilities and Technologies
Key enabling robotics technologies, perception capability and systems integration for recognition, interaction and task execution in complex environments.
National Robotics Programme · Robot Domain Specific
Robot vision, task understanding, environment adaptation, system coordination and validation in real scenarios.
Vision module optimisation and systems integration on Toyota HSR
Led vision algorithm optimisation and systems integration; recognition accuracy for common household items rose from about 80% to about 90%, supporting recognition, picking and handing tasks.
RoboCup / Robocon / RoboRoarz international robotics competitions
Organised and led teams at international robotics events; led vision system development in the RoboCup 2010 domestic service category, where the team placed sixth among 26 international university teams.
Representative results
Foundations of complex hazard recognition, safety-distance estimation and real-robot navigation
Sensors 2023Context-enhanced hazard detection and safe navigation
Key results
- Mean detection accuracy88.71%
- Mean IoU0.77
- Real-robot validationCompleted on BELUGA
What the work demonstrates
- Built a context-enhanced multi-layer fusion detection framework for low-feature, easily occluded hazards such as escalators, stairs and glass doors.
- Hazard-recognition outputs are converted into risk-area representations and safety-distance estimates that further drive robot path adjustment.
- Real-environment navigation was validated on the BELUGA mobile robot, closing the recognition–estimation–planning–execution loop.
Relevance to this project: This foundation transfers low-feature hazard recognition to joint perception of associated objects — the elderly, wheelchairs, walkers, IV stands and hospital beds; extends static safety distance into a dynamic safety envelope combining relative speed, braking capability, prediction uncertainty and object fragility; and continues the real-robot loop where detection results drive navigation behaviour, validating slow-down, wait, detour and stop strategies.
Foundations of non-visual situational awareness, hazard-zone estimation and path replanning
Mathematics 2023Hazard area avoidance via acoustic situational awareness
Key results
- Frame recall92.38%—98.78%
- Sound source localisationValidated
- Coverage path replanningCompleted
What the work demonstrates
- Microphone arrays drive sound-event detection, sound-source direction finding and hazardous-zone estimation.
- The research goes beyond sound classification, converting non-visual perception into spatial risk distributions.
- Risk areas further drive coverage-path replanning, turning multimodal risk cues into executable robot behaviour.
Relevance to this project: It extends single-vision obstacle avoidance into multimodal safety awareness: when the RGB-D camera is occluded, in low light or field-limited, sound events, source directions and environmental anomalies serve as supplementary risk cues. The situational-awareness–risk-region–path-replanning logic writes collision risk and prediction uncertainty into the local cost map and safety-control constraints, giving the safety arbiter multi-source grounds to choose slow-down, wait, detour or stop.
System-integration foundations of perception–planning–execution on the Toyota HSR platform
Visual perception, 3D localisation and coordinated execution for service robots
Key results
- Mean waste detection accuracy96%
- Online door handle detection accuracy91.2%
- System loopPerception–planning–execution
What the work demonstrates
- Tabletop waste recognition, RGB-D 3D localisation, PRM/DFS path planning and manipulator execution.
- Wall and floor semantic segmentation, real-time obstacle detection and wall following.
- Door-handle recognition and coordinated base–manipulator cleaning, connecting vision, 3D coordinate transforms, task planning, base control, manipulator operation and ROS communication.
Relevance to this project: This foundation directly supports the engineering realisation of the edge safety controller, the ROS 2 safety middleware, the wheeled-robot validation prototype and small-scale trials in assisted-care settings. Together the three earlier results form a progression from algorithm to system, from single modality to multimodality, and from risk recognition to robot behaviour adjustment.
LET'S BUILD SOMETHING USEFUL
Your next idea. Made real.
Tell us about your equipment, workflow or learning goals. We will help define the next step.